February 16, 2000
ALCOHOL
Philip J. Cook and Michael J. Moore
Duke University and NBER
Prepared for the Handbook of Health Economics, Joseph P. Newhouse and Anthony Culyer, editors.
Thanks to Michael Grossman, Will Manning, and Harri Sintonen for their very helpful comments. All
errors remain the authors’.
Abstract
Excess drinking is associated with lost productivity, accidents, disability, early death, crime,
neglect of family responsibilities, and personality deterioration. These and related concerns have
justified special restrictions on alcoholic-beverage commerce and consumption. The nature and extent
of government involvement in this arena vary widely over time and place, and are often controversial.
Economists have contributed to the evaluation of alcohol policy through empirical work on the effects of
alcohol-control measures on consumption and its consequences. Economics has also provided an
accounting framework for defining and comparing costs and benefits of interventions, including excise
taxes.
Outside of the policy arena, economists have analyzed alcohol consumption in the context of
stretching the standard model of consumer choice to include intertemporal effects and social influence.
Nonetheless, perhaps the most important contribution by economists has been the repeated
demonstration that there is nothing unusual about alcohol in at least one essential respect: consumers
drink less ethanol (and have fewer alcohol-related problems) when alcohol-beverage prices are
increased.
Important econometric challenges remain, including the search for a satisfactory resolution to the
conflicting results on the effect of price changes on consumption by consumers who tend to drink
heavily. There are also unresolved puzzles about the relationship between drinking and productivity;
even after controlling for a variety of other characteristics, drinkers tend to have higher earnings than
abstainers, and women's earnings (but not men's) tend to increase with alcohol consumption.
Philip J. Cook, Sanford Institute of Public Policy, P.O. Box 90245, Durham, NC 27708
Michael J. Moore, Fuqua School of Business, P.O. Box 90120, Durham, NC 27708
2
I. Introduction
While the production and sale of alcoholic beverages constitute a minor share of national
product in advanced economies, the health-related and socioeconomic consequences of alcohol
consumption are quite severe. Excess drinking is associated with lost productivity, disability, early
death, crime, neglect of family responsibilities, personality deterioration, and other problems. These and
related concerns have long engendered public support for treating alcohol differently from other
commodities. In the private sector, religious teachings and cultural norms are reinforced by rules
restricting drinking in the workplace, schools, and other gathering places. There is also pervasive
government regulation of the production and marketing of alcoholic beverages, coupled with high excise
taxes. The nature and extent of government involvement in this arena vary widely over time and place,
and are often controversial.
Economists have contributed to the evaluation of alcohol policy through empirical research on
the effects of alcohol-control measures on alcohol consumption and its consequences. They have also
made progress in developing and implementing a social-cost-accounting framework for evaluating such
measures that gives due weight to the benefits of drinking, and attempts to distinguish consequences
(both good and ill) that are borne by the drinker and those borne by third parties.
Outside of the policy arena, economists have been interested in studying consumer behavior
with respect to alcohol because of several intriguing characteristics that distinguish it from other
commodities. For one thing, drinking is habit forming; current consumption may have a profound
influence on future tastes. There are other delayed consequences of current drinking as well, ranging
from hangover and accidental injury in the short term, to loss of reputation, reduced earnings, and organ
3
damage over the longer run. Some people develop such a strong taste for alcohol that they are willing to
incur great personal costs for the sake of their habit. The economics of self-control and addiction has
become an active field, engaging both economists and behavioral scientists, with alcohol as a primary
referent.
A second characteristic of drinking behavior is the importance of social context. A primary use
of alcohol is as a social lubricant, and both the availability of drinks and the pleasure of drinking may
depend on the nature of the occasion. The claim that drinking is a socially contagious activity, long
asserted by sociologists, is just beginning to emerge as a focus for economists.
In this chapter we review the economics literature on drinking and its consequences, and
suggest some promising future research directions. The next section describes the trends and patterns
of alcohol consumption within and across populations, and characterizes some of the recent trends.
Sections 3 and 4 then review the theory and estimation of the demand for alcoholic beverages, noting
the special challenges to empirical work in this area and then reviewing findings with respect to the effect
on various measures of consumption of income, own price, price of other intoxicants, and social
influence.
Section 5 reviews findings on how drinking affects health status and socially relevant behaviors.
The focus is on the economists' contributions to estimating the effects of alcohol-control measures on
these outcomes; these contributions are of direct relevance to the policy debate and have also been
important to our understanding of the epidemiology of alcohol-related problems. A particularly
important consequence of drinking in policy debates historically has been its effects on productivity.
The literature in this area is reviewed in Section 6.
Section 7 reviews the normative literature on alcohol-control measures, with particular focus on
4
alcohol-excise taxes. We conclude with some suggestions for future research directions.
2. Trends and Patterns in Alcohol Consumption
Alcoholic beverages include wines, beers, and spirits. Wines are fermented from the sugars in
fruits or berries (most commonly grapes) or other sources (Keller 1978). Beers are fermented from
grains after the starch in them is first converted to sugar. Spirits are distilled from wines or beers. The
alcohol in all these beverages is ethyl alcohol, also known as ethanol. In beers the alcohol content
ranges from 2 to 8 percent; unfortified wines contain between 8 and 14 percent (with 14 being the
upper limit from the fermentation process). Fortified wines, including vermouth, sherry, and port,
contain about 20 percent, which is achieved by the addition of brandy or pure alcohol. Spirits usually
contain between 40 and 50 percent alcohol.
Consumption of alcoholic beverages varies widely across and within populations, and over time.
Here we report some of the more notable patterns, first for national-level data, and then for data on the
distribution among individuals.
2.1 Aggregate data
In the United States, apparent consumption of ethanol per capita (age 14 and over) peaked in
1980 and 1981 at 2.76 gallons after 20 years of steady growth (Williams et al. 1996). This peak
volume, which is equivalent to about two drinks of beer, wine, or spirits per day for every adult, is the
highest recorded in the 20th century. Drinking has declined since then, and was down 20 percent by
1994. The decline has mostly been in the consumption of spirits; as a result, the share of alcohol from
5
spirits declined from 37 to 30 percent of the total, while beer increased its share from 50 to 57 percent
and wine held a constant 13 percent share.
These statistics are estimates based on state excise-tax records and industry reports, and are
subject to error. They take no account of wastage, illicit production and imports, and legal home
production, all of which may affect observed trends as well as levels of consumption. For example, we
would expect the gap between measured and actual consumption to be influenced by excise tax rates,
the real value of which trended downward during the post-War period. Another potential source of
error is the conversion from beverage volume to ethanol volume. In recent years the conversion factors,
which are estimates of the relevant averages for each beverage category, are 4.5 percent for beer, 12.9
percent for wine, and 41.1 percent for spirits (Stinson et al., 1997).
The consumption trends observed in the United States are similar to those in most other
economically advanced countries. Between the Second World War and the 1970s, consumption
increased in almost all countries that were able to offer reasonably accurate statistics, with the largest
growth rates recorded by countries that started from a relatively low level. Hence the trend was
towards narrowing of international differences in ethanol consumption (Sulkunen 1983). Consumption
per capita fell in most OECD countries during the 1980s (Edwards et al. 1994, p. 35).
Based on 1995 data, apparent per capita alcohol consumption in the United States is about the
same as Canada and the United Kingdom and lower than in Western Europe with the exception of
several Nordic countries. At the very top of the world drinking list are Luxembourg, France, Portugal,
Hungary, and Spain where consumption is half again as high as in the U.S.1 The Japanese average
about the same alcohol intake as North Americans, while the Chinese and residents of other Asian
countries are estimated to drink a good deal less, as do Latin Americans.
6
2.2 Individual differences
Per capita consumption statistics conceal the wide variation among individuals in alcohol
consumption. The National Household Survey on Drug Abuse for 1996 estimated that in the United
States, 65 percent of adults had at least one drink during the previous 12 months, while just half drank
during the previous month.2 The prevalence of self-reported drinking decreases in middle age and is
much lower for women than men. For both men and women the prevalence of drinking increases with
education and family income, and is lower for blacks than whites. Survey data also provide information
on how much alcohol is typically consumed by those who do not abstain. The proportion of adults who
"binged" (5 or more drinks on a single occasion) in the previous month follows the same patterns with
respect to age and sex, but with respect to race and education there is little difference across groups.
The distribution of alcohol consumption among drinkers is highly skewed to the right in most
every population that has been studied. The log-normal distribution provides a reasonably good fit to
the empirical histogram (Edwards et al. 1994). Hence consumption is highly concentrated among a
relatively few people. Gerstein (1981) conveys this result with the following image:
If we were to reduce the overall U.S. consumption curve to a representative sample of 10
drinking-age adults, their annual consumption of absolute ethanol would not be very different
from the following rough approximation: 3 nondrinkers, 3 drinking a gallon among them, and the
others drinking 1.5, 3, 6, and 15 gallons respectively (p. 193)."
The heaviest drinkers are of considerable importance to the alcoholic-beverage industry; for example, if
the top decile somehow could be induced to cut their consumption level to that of the next lower group
7
(the ninth decile), then total ethanol sales would fall by over one third.
Actually it is unlikely that any available intervention (short of rigorously enforced rationing)
would truncate the distribution in this fashion. International comparative studies of drinking distributions
have found remarkable consistency in their shape. As per capita consumption by drinkers increases
(over time or across populations), the consumption associated with each quantile tends to increase in
proportion (Edwards et al. 1994, 83-90). Hence, we would expect that as per capita consumption
declines, as it has in many Western countries in recent years, then the prevalence of heavy drinking
(defined by any absolute standard) will decline also -- indeed, given that the downward shift is
proportional at all levels, the prevalence of heavy drinking will decline more rapidly than average
consumption. This possibility is compatible with trends in the adverse consequences of heavy drinking
since the drinking downturn in the early 1980's (see Section 5).
2.3 Expenditures
Retail sales of alcoholic beverages in the United States totaled $99 billion in 1996, equally
divided between on-premise and off-premise (Adams Business Media 1997). That represents just 2
percent of personal consumption expenditures in that year, although in fact not all sales are paid for as
household consumption items -- by one estimate, employers pick up the tab for about 20 percent of
sales (Sammartino 1990, p. 76). While the division of expenditures between on-premise and off is 5050, most alcohol by volume is consumed off-premise: 79 percent of spirits and wine, and 76 percent of
beer, in 1996 (Adams Business Media 1997).
While these details differ from place to place, it is true throughout the developed world that the
alcoholic beverage industry is relatively minor in terms of direct economic importance. But the health
8
and social consequences of drinking amplify its importance.
3. Framework
Economists who study alcoholic beverages have focused on those qualities that are distinctive
and important about this class of commodities. First, alcohol is an intoxicant; consumed in sufficient
quantity in a single session it impairs mental and physical functioning and is potentially toxic. Second,
alcohol consumption has direct intertemporal consequences; past consumption affects tastes for future
consumption in a predictable way (habit formation), and chronic alcohol use affects physical and mental
health over the course of years or decades. In the aggregate, drinking has important economic effects
through its influence on the public health and safety, productivity, family functioning, criminal activity, and
so forth. These special features define much of the research agenda for economists and other social
scientists.
In particular, the dependent variables used in demand studies are not only measures of the
overall quantity per unit of time, but also measures of the timing and circumstances of consumption and
its consequences. The determinants of quantity demanded include not only prices and income, but also
measures of past and anticipated future consumption, and the availability of other intoxicants.
The research agenda has also been shaped by the special features of alcoholic-beverage
markets. Commerce in alcoholic beverages is subject to extensive government regulation, so quantity
demanded is mediated not just by price but a variety of other alcohol-control measures. Indeed, at
different times and places the manufacture or trade in all or at least some types of alcohol has been
prohibited. Short of prohibition, common legal restrictions specify when, where, how, and to whom
alcohol may be sold. These restrictions include government monopolization of manufacture or trade, or
9
if private vendors are permitted, special licensing requirements. Also common are limits on advertising
and hours of operation, prohibition on sale to minors or drunks, restriction on the content and labeling of
beverages, and quantity rationing.3 All such control measures are potentially grist for the research mill.
A schematic diagram of the various relationships that comprise the economic literature on
alcohol is presented in the figure below. In this diagram, "current drinking" patterns (both on and off the
job) may affect productivity, which in turn determines individual earnings. The diagram also depicts an
indirect effect via the influence of drinking history on the accumulation of human and health capital. The
final link represents "reverse causation," in which current consumption is affected by earnings.
Figure 1 about here
To fix ideas for our review, we specify a simple structural model of alcohol consumption and its
consequences that incorporates some of these considerations. The model specifies intertemporal
demand imbedded in a health-production process of the sort developed by
Grossman (1972).
We specify the demand for alcohol in the rational-addiction form developed in seminal papers
by Becker and Murphy (1988) and by Becker, Grossman and Murphy (1991). Prior empirical and
theoretical research developed the notion of intertemporal dependency via models of habit formation, in
which current utility depends on past as well as current consumption of the addictive good. These habitformation models implicitly assumed that the consumer was myopic, in the sense that he ignored the
possibility that his current choices would affect future tastes or health. A "rational" consumer, on the
other hand, would recognize the future consequences of current drinking decisions and take them into
10
account in planning a sequence of consumption levels. Current demand then becomes a function of past
consumption, current prices and income, and expectations concerning future drinking and prices.
Actually, the future may loom more broadly in affecting current drinking than is implied by the
standard rational-addiction model. Given the habit-forming nature of drinking, and its other
consequences, it is reasonable that some consumers will moderate their drinking in response to
expectations concerning the effect of current consumption on future schooling, employment, family, and
health status. We will ignore these influences in presenting the model, but return to them in the
discussion.
To capture the potential for habituation, let current-period utility be a function of health, other
goods, and current and past consumption of alcohol. Thus, the period-specific utility function is
Ut=U(Ht,Xt,At,At-1), where Ht denotes health capital in period t, Xt other goods, and At the
consumption of alcohol..
Let the current wage be denoted by Wt, labor hours by Lt, and non-labor income by It. Income
in any period, Yt, is then WtLt+It. For simplicity, assume full depreciation of the addictive stock in each
period, with an individual discount factor β =1/(1+r), where r is the discount rate. The individual
chooses to maximize lifetime utility, V, which is the discounted sum of the period-specific utilities. That
is,
∞
V =∑β
t −1
U
t
t =1
(1)
One "outcome function" will be treated here: a health-production function.4 This is a
technological relationship, in the sense that variation in the health output is produced by variation in the
11
underlying input of alcohol. Alcohol may impair the ability to perform various tasks, such as driving a car
or operating a machine, or it may damage health directly, as in the case of some heart diseases,
cirrhosis, high blood pressure, and hemorraghic strokes. On the other hand, moderate alcohol
consumption appears to have beneficial effects, helping prevent certain types of heart disease.
Health is produced using medical care, M, and A as inputs, conditional on the preexisting
health-capital stock and on other determinants of health, vH
(2)
Ht = H(Mt,At,Ht-1,vH).
The effect of M on H is assumed positive; the marginal effect of A may depend on the level of
A; at sufficiently high levels the marginal effect is surely negative.
In general, the assumed shape of the production relationship given by (2) is crucial in
determining the validity of empirical estimates, and in inferences made about the efficacy of policy
interventions aimed at reducing the harmful effects of alcohol consumption. We will expand on this issue
in a subsequent section.
The individual has a unit of time available each period to spend working, L, or relaxing, R. The
time constraint is then
(3)
L + R = 1.
Under certain assumptions, and if the individual maximizes (1) subject to the constraints given by
(2), (3) and the intertemporal budget constraint, we can derive linear demand functions (or more
precisely, equations of motion) for A and M, and a linear labor supply function. In particular, if (1) and
(2) are quadratic, and holding the marginal utility of wealth constant, we have:
(4)
At=α 0+α 1At+1+α 2PAt+α 3At-1+α 4Wt+α 5Pt+α 6Zt+ε At.
(5)
Mt=δ 0+δ 1At+1+δ 2PAt+δ 3At-1+δ 4Wt+δ 5PMt+δ 6Zt+ε Mt.
12
and
(6)
Lt=γ0+γ1At+1+γ2PAt+γ3At-1+γ4Wt+γ5PMt+γ6Zt+ε Lt.
where PA and PM represent the prices of alcohol and medical care respectively, and Z represents one
or more covariates.
The interior solution represented here is clearly more appropriate for aggregate data, as zero
values are a common feature of microeconomic data on alcohol consumption, medical care, and labor
supply. The Kuhn-Tucker approach to modeling commodity demands for individual-level data is
developed in Wales and Woodland (1983) and in Pudney (1989).5
In addition to these equations of motion, we could also estimate the production relationship
given by the health-production function in (2). In the model developed above, this was assumed to be
quadratic and, to be fully consistent with that model, a quadratic version of (2) could be estimated.
Most estimates of (2) that have appeared in the literature have been linearized versions of these
production relationships. Health-production functions have been the focus of the epidemiological
literatures on the health effects of drinking, where health is measured variously by mortality, morbidity,
and indicators of physical or mental health. In some applications, health is proxied by the use of medical
inputs as well. The measures of alcohol consumption in these health-production productions in some
studies allow for non-linearities, either through the use of dummy variables relating to intensity of use, or
with quadratic terms in quantity of alcohol consumed.
The typical health production relationship estimated in the literature, however, has the form:
(7)
Ht=h0+h1At+h2Mt+h3Y t+h4Zt+ε Ht.
13
4. Demand for Alcoholic Beverages
Econometric studies of demand for alcoholic beverages have been conducted with a wide
variety of data sets. Edwards et al. (1994) tabulate results from such studies for 18 countries.
Estimated elasticities for beer, wine, and spirits differ widely over time, place, data set, and estimation
method, but one conclusion stands out: In almost every case the own-price elasticities are negative. In
that fundamental respect, at least, it appears that alcoholic beverages are like other commodities.
Clements et al. (1997) report results for their estimates of systems of demand equations for
Australia, Canada, Finland, New Zealand, Norway, Sweden, and the U.K., in each case using
aggregate data covering about 30 years. Their average own-price elasticities are -.35 for beer, -.68 for
wine, and -.98 for spirits. Beer has the lowest own-price elasticity of the three beverage types in all 7
countries. This pattern is well established though still somewhat mysterious.6
The focus of these studies is on average (per capita) or total quantity, standard fare for empirical
analysis of markets. However, since our interest here is on health effects, the average is of less interest
than other measures of a population's consumption of alcohol, including the prevalence of abstention, of
bingeing, and of chronic heavy drinking. Distinctions by age, sex, and drinking history of the consumer
are also relevant to consequences. Analysis of this sort cannot be performed with aggregate data, but
rather requires data on individuals. What follows, then, is a review of studies that utilize survey data on
individual respondents, or, in a few cases, experimental data.
4.1 Measurement issues
There are a number of surveys based on U.S. national samples that include items on alcohol
14
consumption (Sindelar 1993). Given the importance of intertemporal influences in analyzing demand,
panel data are particularly useful. In many respects the most complete data are provided by the
National Longitudinal Survey of Youth (NLSY), which has interviewed its initial sample of over 12,000
youths every year since 1979, at which time they were aged 14-21. The questionnaires included items
on alcohol consumption from 1982-85 and again in more recent years, and also have included standard
inventories on alcohol dependence and abuse. Over the years a wealth of other information has been
collected on parents and siblings, schooling, work, earnings and other income, use of other intoxicants
and tobacco, health status, family formation, and other matters. Price information is not obtained from
the respondents, but can be imputed from knowledge of the respondent's place of residence. Of course
the NLSY data are limited to a particular set of cohorts. The literature includes results from the NLSY,
but from a variety of other surveys as well.
The drinking items in these surveys are of suspect quality. Self-reports of alcohol consumption
tend to understate actual consumption. In general population surveys, comparisons of self-reported
drinking with sales data suggest that such surveys typically capture only 40-60 percent of actual
consumption (Midanik 1982). In addition to the obvious explanation, that respondents tend to
underreport their drinking, it is also likely that national household surveys have a somewhat biased
sample. Some heavy-drinking groups are underrepresented either because they are not in the sampling
frame (e.g., homeless people and those in institutions) or because it is difficult to contact them and gain
their cooperation (e.g., people with transitory life styles) (Polich and Orvis 1979, p. 56).
With surveys, the details matter, as illustrated by a comparison of two sets of survey results for
high-school seniors in 1982. We compared NLSY data with data from an ongoing survey of cohorts of
high-school students called Monitoring the Future (MTF), finding that MTF generated much higher
15
estimates of drinking prevalence and of binge drinking than NLSY (Cook, Moore and Pacula 1993).
One likely explanation is that MTF questionnaires were administered in the classroom, while NLSY
respondents were interviewed at home. In subsequent years the drinking results for these two samples
of respondents converge.7
The consequences of response error for econometric results depend on its structure across
respondents and over time. In fact there is little direct evidence on whether response errors tend to be
additive or proportional, random or systematic. It does seem reasonable to view self-reported
consumption by a respondent as a lower bound on true consumption; those who report bingeing really
do, and so do some of those who claim to be abstainers or light drinkers. But there is little direct
evidence on this matter.8
The implications of measurement error for estimates of equations (4)-(7) are far from clear. If
the drinking variable is used as an explanatory variable, bias will be introduced into the estimated effect
of this variable, and into the estimated coefficients of the remaining explanatory variables. If the drinking
variable is the dependent variable, and measurement errors are correlated with quantity consumed or
with explanatory variables, then the coefficient estimates will be biased as well.
The measurement-error problems become particularly acute in the microeconometric literature
on alcohol demand. Since the dependent variables of interest in this literature are typically discrete or
bounded, nonlinear estimation techniques are usually employed. Even random errors in the dependent
variable impart bias to estimated coefficients. The importance of individual heterogeneity in alcohol
demand and human- or health-capital models has led some researchers to resort to fixed-effect-type
models using panels of individual data. Nonlinear fixed-effect models, which are notoriously difficult to
estimate (Chamberlain 1984), become even more so in the presence of measurement errors.
16
Data problems also arise in the choice of an indicator of price. In any one market, the price of a
drink differs widely depending on the type of beverage, the brand, the type of retailer, and whether the
purchase is for consumption on-premise or off (Grunewald et al. 1996). What is needed for crosssection or intertemporal comparisons is a price index of some sort. In the absence of a local-area
index, most investigators have used either the state excise tax rate (usually just on beer), or the average
price of a 6-pack of a popular brand of beer (available for each of a number of urban areas from data
compiled by the American Chamber of Commerce Researchers Association, or ACCRA), sometimes
adjusted for an index of the local cost of living. This ACCRA price estimate is in principle a more
sensitive indicator of local market conditions, but is of course measured with greater error than the
excise tax rate and is not computed for all areas of residence.
4.2 Drinking by Youths
Much of the econometric research has focused on drinking by youths. Teenagers and young
adults are of special concern for several reasons. First, youths exhibit relatively high rates (compared
with their elders) of binge drinking and involvement in motor-vehicle accidents and violent crime
(Grossman et al. 1994). Second, to the extent that drinking is habit forming, youthful drinking sets the
pattern for later consumption. And third, drinking behavior during the transition from adolescence to
adulthood may have important consequences for human capital and family formation (Cook and Moore
1993).
Beginning with Grossman, et al. (1987) and Coate and Grossman (1988), a series of studies
have documented the sensitivity of youthful drinking to both minimum- drinking-age (MDA) laws and to
beer prices. (The focus on beer is dictated by the fact that most ethanol consumed by youths in the
17
U.S. is in the form of beer.) A number of studies using five different data sets suggest that how often
youths drink and how often they binge are both importantly related to price and MDA (Grossman et al.
1994). An exception is Chaloupka and Wechsler (1996), that analyzes drinking by college students
using a large one-time survey. They found that the price of beer has no discernible effect on drinking
practices of male students. A possible explanation, they suggest, is that much of their alcohol
consumption is in group settings where individuals do not pay by the drink. Price also has a significant
effect on the likelihood of abstention, though ironically this is not true for the MDA (Moore and Cook
1995).
Evidence that drinking is habit forming for youths comes from analyses of panel data. Moore
and Cook (1995), for example, analyze four waves of NLSY data. The likelihood of drinking in 1985
is related to whether or not the respondent reported drinking in previous years according to the
following equation:
Pr[d85 = 1] = .16 + .36d84 + .23d83 + .16d82
Thus past drinking predicts future drinking, and the recent past predicts more closely than the distant
past. Of course, this result does not prove habit formation. It may in part reflect persistent differences
among youths with respect to their taste for alcohol (or their circumstances). Distinguishing between
state dependence (habit formation) and persistent heterogeneity is a common problem in labor
economics and other areas of application (Heckman 1981). One approach for ruling out the
heterogeneity explanation is through use of instrumental variables for past drinking. We took the
somewhat more direct approach of estimating the effect of alcohol availability (as indicated by excise tax
and MDA) at age 14 on drinking a few years later. The results confirm the habit-formation explanation
18
for both drinking and bingeing.
The rational-addiction framework described in Section 3 suggests that current drinking choices
will be influenced by the future as well as the past and present. Youths who anticipate higher alcohol
prices in the future, or (perhaps more important) foresee circumstances such as childbearing in which
drinking will have increased nonmonetary costs, will curtail current drinking if they believe that their
drinking is habit forming. Demand estimation in the presence of rational addiction requires indicators of
anticipated prices and nonmonetary costs. Empirical implementation of this framework has proven
difficult in practice (Chaloupka and Grossman, 1994; Moore and Cook 1995; Grossman et al. 1998).
So far our review has been concerned with a single intoxicant, alcohol. Other intoxicants,
especially marijuana but other illicit drugs as well, are widely available to youths in the U.S. It is of no
small concern whether the MDA and other alcohol-control measures may lead youths to substitute illicit
drugs for alcohol.
The 30-day-prevalence results from the NLSY data in 1984 suggest to the contrary that
drinking and marijuana use go together: 27 percent of drinkers, but only 5 percent of abstainers,
reported using marijuana (Pacula 1998). But this pattern of multi-drug use, while suggestive, does not
demonstrate that alcohol and marijuana are complements. As before, it may simply reflect heterogeneity
in the taste for intoxicants. More direct tests have had mixed results. Pacula (1998) reports that higher
beer prices significantly reduce the demand for both alcohol and marijuana, suggesting
contemporaneous complementarity. Chaloupka and Laixuthai (1997) find, on the other hand, that an
increase in the full price of marijuana (indicated by the stringency of state laws and a measure of retail
price) increases drinking and bingeing, suggesting that alcohol and marijuana are substitutes.
This issue is resolved to some extent by studies of the effect of alcohol-control measures on
19
highway fatalities reviewed in Section 5. The indirect effect (via marijuana use) of such measures on
intoxicated driving is implicit in these results. The evidence that raising the MDA reduces youth
involvement in fatal crashes suggests that substitution to marijuana (if any) is either relatively small or
benign.
4.3 Heavy Drinking
Some people acquire such a strong taste for alcohol that they are willing to sacrifice their health
and much else for the sake of continued heavy drinking. Symptoms of strong commitment to alcohol
are the basis for a diagnosis of alcohol dependence, a form of mental illness more commonly known as
alcoholism. It is a widespread presumption that alcoholics will do whatever is necessary to maintain
their drinking at a high level, including substituting cheaper sources of ethanol or cutting back on other
living expenses. In particular, an increase in the excise tax would likely make their lives more difficult
but not cause them to cut back.
Experimental evidence, however, suggests that this image is not entirely correct (Babor 1985).
In an early experimental study, Mello et al. (1968) compared drinking patterns of 14 male alcoholics as
a function of the cost of a drink. Subjects who were required to work twice as hard for their alcohol
drank half as much as comparable subjects in an identical situation. Other experiments with inpatient
alcoholics found that their drinking could be reduced by contingent loss of privileges and financial
incentives for abstinence (Mello 1972; Babor et al. 1978). Another study recruited 34 adult males from
the community for an experiment comparing the responsiveness of the 20 casual drinkers and the 14
heavy drinkers to changes in price. The response of both groups to a "happy hour" in which prices
were cut in half was to approximately double the number of drinks they consumed (Babor et al. 1978).
20
The evidence from outside the laboratory is primarily from studies that use the cirrhosis mortality
rate as an indicator of the prevalence of chronic heavy drinking. These studies generally confirm the
experimental results; they are reviewed in detail in Section 5.
A different conclusion emerges from the application of quantile regression analysis to survey
data. Manning et al. (1995) used data from the National Health Interview Survey to analyze how price
elasticity changes with drinking level. They find that the decision to drink is responsive to price level,
and that among those who report drinking, the estimated price elasticity follows a U-shaped pattern with
respect to relative consumption. The elasticity peaks at -1.2 at the median, and approaches zero at
higher levels. At the highest level of consumption, representing the 95th percentile, the estimated
elasticity is slightly positive. This result appears to confirm the conventional wisdom that heavy drinkers
will find a way to maintain their drinking in the face of modest changes in price. The apparent
contradiction between this result and other findings has not been explained.
4.4 Social Influence
While the decision of how much to drink depends to an extent on individual tastes and financial
circumstances, it is also true that drinking is a social activity. The utility of taking a drink at a particular
time and place depends not only on individual tastes but also on the social setting; for many people,
drinking with others is more enjoyable than drinking alone, and associating with those who are not
drinking, or who disapprove of drinking, may lead one to substitute another type of beverage. Of
course the social setting may also influence drinking decisions directly, by determining the availability of
drinks.9 Given these mechanisms of social influence, it seems reasonable to presume that individual
drinking is influenced by the “wetness” of the social environment.
21
If individual drinking decisions are positively linked to the drinking practices of others, then there
will be a “social multiplier” in the response of aggregate alcohol consumption with respect to prices,
income, and other external influences; social influence will amplify the direct effects of such variables.
This mechanism may be particularly important for initiation into drinking. Indeed, the assumption that
peers are central to adolescent alcohol and drug use is reflected in the social-influence paradigm
underlying many prevention programs (Bauman and Ennett 1996).
One type of evidence in support of this view is that adolescents whose friends drink are far
more likely to drink themselves. For example, Norton et al. (in press) studied drinking in 36 schools,
finding that adolescents in schools with a high prevalence of drinking were more likely to drink
themselves. This result holds after controlling for various individual, household, and neighborhood
characteristics. Indeed, the estimated effect is very large, suggesting that an increase of 10 percentage
points in group drinking is associated with an equal increase in the likelihood of individual drinking.10
But this result is compatible with several other mechanisms besides social influence (Manski
1995): First, it may be the result of an endogenous selection process where some parents consider the
behavior of the local adolescents in deciding where to live. Second, it may reflect a “contextual” effect,
where the individual’s drinking behavior is influenced by other characteristics of the group (commitment
to getting a good education) but not by their drinking per se. Third, it may be true that youths within the
same group share some important but unobserved aspect of the environment, such as whether local
merchants are willing to sell alcohol to youths.
In the absence of experimental data, the identification problem here is severe. Instrumentalvariables methods may provide estimates of the extent to which an individual’s drinking is influenced by
the group without demonstrating that it is the group’s drinking (rather than some correlate of their
22
drinking) that is the direct cause (Norton et al., in press). This distinction is important because if it is not
the group’s drinking per se, but rather some other mechanism of social influence, then there is no social
multiplier (Gaviria and Raphael, in press). For example, if higher-education plans influence drinking, and
youths are influenced by their school peers in evaluating the prospect of higher education, then a student
attending a school where most students have no plans for higher education will, like her peers, tend to
drink more than otherwise, even if drinking is not contagious.
But assuming that drinking is subject to a positive-feedback effect through the group, the result
is to increase the elasticity of market demand with respect to own price and other determinants.
Further, for a given price regime, aggregate quantity consumed may depend on the extent of social
mixing among individuals with different drinking propensities, a factor which may be deemed relevant,
for example, in setting housing rules on college campuses.
4.5 Advertising
Individuals learn about alcoholic beverages and drinking not only from friends, but from a
variety of other sources as well, including the popular media, church, the classroom, consultations with
health-care providers, labels on beverage containers, and so forth. From this array of potential
influences on the demand for alcohol, the economists’ contribution has been largely confined to the
analysis of commercial advertising. The qualitative issues here are much the same as in the case of
tobacco, and we refer readers to the excellent discussion of advertising in that chapter of this volume.
Producers sponsor ads in order to increase the demand for their particular brands, and it is
possible that the cumulative effect has more to do with the distribution of market share than with overall
quantity of drinking. Econometric studies of commercial advertising have reported differing results with
23
respect to the estimated effect on consumption, as reviewed in Saffer (1995). Saffer (1997) argues that
we would expect the effects of advertising to be subject to diminishing returns, and that rivalry among
producers may yield investments in advertising that at the margin have little effect on consumption. Thus
data from a regime where advertising is unconstrained provide information on the (possibly negligible)
marginal effects, but not on the overall effects, of advertising. That may help explain the null results
reported by Nelson and Moran (1995), in their study of U.S. national time-series data, and of Gius
(1996), in his study of brand advertising for 15 brands of distilled spirits; these and a number of earlier
studies report a negligible effect of overall advertising on aggregate consumption.
Saffer (1991) sought to measure infra-marginal effects of advertising by analyzing national ad
bans. His study was based on a time series of cross-section data on 17 countries over 14 years. He
found that after controlling for price and other factors, a ban on spirits advertising was associated with a
16 percent reduction in ethanol consumption, and that a ban on advertising of all types of alcohol
lowered consumption by an additional 11 percent. These results can be challenged on the basis that the
“assignment” of ad bans to countries is not necessarily exogenous. It may also be true that advertising
influences prices, as would be true if advertising strengthens brand loyalty and thus reduces own-price
elasticity. In this case the “ban” coefficient provides a biased estimate of the full effect.
Public concern about alcohol advertising is more focused on alcohol abuse than on average
consumption levels, and so it is of particular interest to measure the effect of commercial advertising on
such outcome measures as highway fatalities and cirrhosis mortality (Saffer 1991). For example, a
study by Saffer (1997) related alcohol advertising messages to traffic fatalities, utilizing panel data on the
75 top television markets. This study allowed for the possibility that advertising is endogenously linked
to drinking (and hence to alcohol-related crashes). The results suggested that a ban on broadcast
24
advertising of beer and wine in the U.S. would reduce traffic fatalities by about 5 percent.
At present the methodological difficulties in studying the effects of commercial advertising on
alcohol consumption and abuse have precluded a confident conclusion about whether the regulation of
commercial advertising is a potentially important policy instrument.
5. Consequences of Alcohol Consumption and Taxation
The health and social consequences of drinking render it an important problem for nations
where alcohol consumption is common, and an incipient problem for others. Alcohol-related health
problems include both the short-term consequences of intoxication and the long-term consequences of
chronic heavy drinking. The former category includes injuries sustained in traffic crashes, industrial
accidents, drownings, and alcohol overdose, as well as from intentional violence. The latter, chronic
category includes damage to the liver and other organs, impaired cognition and immune-system function,
and alcohol dependence. Alcohol is also teratogenic; drinking during gestation may cause persistent
abnormalities in physical and cognitive development of the child (Larkby and Day 1997).
The U.S. Centers for Disease Control (CDC) estimated "alcohol-related mortality" for 1987 to
be 105 thousand, 4.9 percent of all deaths in that year (CDC 1990). By this estimate, alcohol is a
smaller problem than smoking but a far larger problem than drug abuse. Nearly half (46 percent) of the
"alcohol problem" in the United States stemmed from accidents and intentional violence. By another
conventional public-health measure, "years of life lost before age 65," violent deaths constitute fully 80
percent of the alcohol problem in the United States (CDC 1990). Thus the chronic effects of excess
consumption have a relatively small effect on life expectancy.
25
A number of other countries have still higher death rates from heavy drinking, particularly in
central Europe and other nations where per capita consumption is high. Perhaps most notable in this
regard is the sharp drop in Russian male life expectancy during the 1990s, which has been attributed in
part to the surge in alcohol consumption following the end of the Gorbachev reforms. From 1990 to
1994 male life expectancy at birth fell by 6.2 years to 57.6. One analysis concluded that "Although
factors such as nutrition and health services may be involved, the evidence is that substantial changes in
alcohol consumption over the period could plausibly explain the main features of the mortality
fluctuations observed (Leon et al. 1997, p. 383)."
The health consequences of drinking alcoholic beverages are apparently not entirely negative.
Historically and even in modern times and places where the only available water for drinking is
contaminated, alcoholic beverages have offered a nearly germ-free alternative (Vallee 1998). And the
accumulating evidence suggests that moderate drinking protects against coronary heart disease.
In this section, we review key studies on the relationship between alcohol consumption, alcoholcontrol measures, and several health outcomes: motor-vehicle fatalities, cirrhosis mortality, medical care,
heart disease, and crime.
5.1. Motor-vehicle mortality
Several of the U.S. studies of alcohol-control policies and highway safety are summarized in
Table 1. The majority of these studies utilize state-year panel data on fatalities, coupled with indicators
of the minimum drinking age (MDA) and price or tax changes. Per capita income is typically utilized as
a control variable, along with fixed effects for state and year, and in some cases autocorrelation
corrections.11 Most of the state-level studies, with the exception of Males (1986), conclude that
26
highway fatalities decline when the minimum purchase age or alcohol excise taxes are increased. (A null
finding on price is reported in Sloan et al. (1994)). Ruhm (1996) suggests that changes in alcoholcontrol measures may be confounded with such omitted variables as grassroots campaigns against
drunk driving and state economic conditions.
This "reduced form" approach to studying highway fatalities and other remote effects of alcoholcontrol measures requires some motivation (Cook 1981). As described in Section 3 above, prices and
other control measures influence outcome measures, if at all, through their effect on drinking. Consider
the following links:
Link 1: Increased excise taxes on alcoholic beverages reduce per capita consumption of
ethanol;
Link 2: A reduction in average consumption of ethanol is associated with reduced prevalence of
intoxication;
Link 3: A reduction in intoxication prevalence is associated with a reduced prevalence of driving
under the influence (DUI);
Link 4: A reduced DUI prevalence reduces the motor-vehicle-fatality rate.
One approach to estimating the effect of excise taxes on motor-vehicle fatalities is to estimate each of
these structural relationships, as in equations (4) and (7). But even if we were able to do so, the result
would be less persuasive than the "reduced form" estimate. For one thing, the intermediate variables,
particularly the measures of drinking, are subject to large errors in measurement. For another, these
intermediate variables are not precise enough to ensure that the links join into a single chain of argument.
It is possible, for example, that excise taxes reduce the frequency of intoxication at home but not away
27
from home, in which case the excise tax would not much affect DUI prevalence even though links 1, 2,
and 3 are each generally true.
Since first reported by Cook (1981) and Cook and Tauchen (1982), the reduced-form
estimates have become standard practice in exploring the effects of alcohol-control measures on a
variety of outcome measures. An ancillary benefit of this approach has been to help establish the causal
importance of drinking with respect to certain outcomes. For example, social scientists have tended to
favor explanations for the observed association between crime and drinking that deny the direct causal
influence of drinking (Collins 1989). But evidence that higher alcohol prices reduce some kinds of crime
suggests that alcohol is the culprit after all.
Returning to our review of drinking and motor-vehicle crashes, we see that two of the studies in
Table 1 utilize microdata on risk-taking behaviors to examine the effects of availability restrictions on
drunk driving. Kenkel (1993) analyzes the Health Promotion and Disease Supplement to the 1985
Health Interview Survey, which contained information on drinking and driving practices. He
incorporates alcohol-control measures, measures of the legal threat to drunk driving, and an indicator of
health knowledge based on individual awareness of connections between drinking and health risks.
Binge drinking is defined as the number of days in the past year with 5 or more drinks, and drunk
driving by responses to a survey question asking how many times in the past year the sample member
drove after having too much to drink.
Kenkel concludes that increases in price and health knowledge reduce the prevalence of binge
drinking, while a state monopoly in liquor sales is associated with an increase in heavy drinking. Binge
drinking in turn increases the prevalence of drunk driving. Interestingly, the legal-threat variables tend to
reduce drunk driving as well, and by the same mechanism -- the threat of punishment reduces binge
28
drinking rather than persuading people to separate their drinking and driving.
Sloan et al. (1995) also analyze microdata, in this case from the Behavioral Risk Factor
Surveys. In addition to the alcohol-control and deterrence effects, they incorporate indicators of the
tort-liability rules to the binge drinking and drunk driving models. While some effects of the criminal and
legal variables have the expected effects on the two outcome variables, most are not statistically
significant. The price and MDA variables exert a strong effect on drunk driving, primarily through their
effect on binge drinking. Also interesting are the results on the incentive effects of compulsory-insurance
laws and experience rating. It appears that these policies, which tend to raise the price of careless
behavior, lead to significant declines in binge drinking.
5.2 Cirrhosis
Cirrhosis is characterized by a progressive replacement of healthy liver tissue with scarring,
leading to liver failure and death. While it has a variety of causes, alcohol accounts for a majority of
cases within population groups where drinking is widespread; indeed, the cirrhosis- mortality rate has
long been used as an indicator of the prevalence of alcoholism in a population (Bruun et al. 1975). The
likelihood of cirrhosis is closely related to lifetime consumption: according to one review, an individual
weighing 150 pounds who drank 21 ounces of 86 proof whiskey per day for 20 years would have a 50
percent chance of suffering from cirrhosis (Lelbach 1974).
Following a long term increase from 1950 to 1973, cirrhosis-death rates in the U.S. have
declined steadily from a peak of 14.9 per 100,000 to 8.1 in 1992 (DeBakey et al., 1995). There is
substantial variation in cirrhosis-death rates across countries. As reported by Edwards, et al., (1994),
death rates from cirrhosis per 100,000 living in Europe ranged from a high of 54.8 in Hungary in 1991
29
to a low of 2.9 in Ireland. In the countries of the western hemisphere, cirrhosis mortality rates are
highest in Mexico and Chile (48.6 and 46.2), and lowest in Canada (9.3).
There is considerable evidence that cirrhosis-death rates are sensitive to alcohol availability,
suggesting that the group at risk for alcohol-related cirrhosis, long-term heavy drinkers, is at least
somewhat price sensitive. Notable cases occurred during and after the First World War. Alcohol was
diverted to military purposes during the war, and several countries instituted prohibition. The results
have been analyzed in Canada, Finland, and the United States, and in each case the reduction in
availability was accompanied by a rapid and substantial drop in the cirrhosis-death rate. In France,
which ordinarily has a high cirrhosis mortality rate, sharp drops in availability occurred during both
World Wars with particularly dramatic effects on the mortality trend (Bruun et al.1975, 43).12
Cirrhosis mortality is also responsive to small changes in price. Cook (1981) and Cook and
Tauchen (1982), in a longitudinal study of state cirrhosis mortality rates, find that increases in state
liquor-excise taxes lead to an immediate (and statistically significant) reduction.13 While this disease
takes years to develop, death rates respond quickly because the progression of the disease (towards
death) is slowed when drinking is curtailed. Over the long run a reduction in heavy drinking will reduce
cirrhosis mortality still further, since the rate of initiation of cirrhosis will be reduced.
5.3 Medical Care
While it may be true that changes in alcohol-control measures affect medical-care utilization,
there has been no analysis of this linkage. There have been several important studies of the relationship
between drinking and medical-care use, however.
Manning et al. (1989, 1991) examine the effects of heavy drinking on outpatient and inpatient
30
care using data from the RAND Health Insurance Experiment (HIE) and the National Health Interview
Survey of 1983. Indicators of drinking used as predictors of health-care utilization include monthly
volume of alcohol consumed together with dummy variables for former drinker and abstainer. Four
medical-care-utilization variables are analyzed, defined by whether they count inpatient or outpatient
visits, and whether they count all such visits or only those in which the diagnosis was likely to be
alcohol-related. Table 2 summarizes the results for the more comprehensive measure of care that is not
limited to alcohol-related diagnoses. (The results for the alcohol-related diagnoses are very similar.)
The two data sets yield the same results for inpatient care. Former drinkers and abstainers use
significantly more medical care than current drinkers. Among current drinkers, there is no discernible
relationship between amount consumed and frequency of inpatient care. For outpatient care, on the
other hand, the two data sets yield somewhat conflicting results.
One hypothesis to which the inpatient results conform is that abstinence and cessation of
drinking may reflect some underlying health condition that is also associated with medical-resource use.
In the model developed in Section 3, if there are correlations among the unobservables in the regression
equations for health, medical care, and drinking equations, an exogenous adverse health shock might
both increase the demand for medical care and reduce the demand for alcohol. In this instance,
estimation of a model in which medical care is regressed on alcohol use could show a positive
relationship between drinking cessation and the use of medical care that is not entirely causal.14
5.4
Heart Disease
Interest in the beneficial effects of moderate alcohol consumption on the heart was stimulated by
publicity surrounding the so-called "French Paradox," that heart disease is lower among the French than
31
for a number of peoples (including Americans) despite the French penchant for smoking and enjoying
artery-clogging diets. The primary hypothesis advanced to explain this phenomenon is that alcohol
consumption, which is also heavier among the French than in the U.S., is somehow responsible.
Most of what we know about alcohol and the heart comes from prospective studies reported in
the epidemiological and medical literature.15 These studies have consistently reported beneficial effects
of moderate drinking, including the Honolulu Heart Study (Yano et al. 1977), the Nurse's Health Study
(Stampfer et al. 1988), the lipid research clinics follow-up study (Criqui et al. 1987), the British
Regional Heart Study (Shaper et al. 1988), the Kaiser-Permanente Study (Klatsky et al. 1990), and the
Physicians’ Health Study (Camargo et al. 1997). The typical result is that a U-shaped relationship
between drinking and the risk of coronary heart disease (CHD) is found, with the beneficial effect
maximized at about 2-3 drinks per day. Estimates of the beneficial impact range from 25 to 50 percent
reductions in CHD mortality for moderate drinkers.
Of course these nonexperimental associations may have other explanations. Reverse causation
is certainly a problem. Many nondrinkers are former drinkers who have quit for health reasons, and it
should not be surprising that they are more likely to exhibit heart problems than moderate drinkers. But
that is not the whole answer since the relationship persists when those with prior conditions are excluded
from the study. What we do not know is if there are unobserved "third causes" which explain both the
propensity to abstain and the propensity to heart disease. In the next section we discuss the findings on
drinking and productivity, which also exhibit an unexplained penalty for abstainers.
Drinking has other effects on the circulatory system as well. It has been found to reduce
clotting, thus reducing the risk of arterial blockages and ischemic strokes. Through this mechanism,
moderate alcohol use can have an immediate effect on mortality (Dufour 1996). At the same time, this
32
thinning effect increases the likelihood of hemorrhagic stroke. Hypertension, one of the primary risk
factors in heart disease, is also increased by regular alcohol consumption. Finally, rhythm disturbances
leading to heart attack are also more likely following spells of heavy drinking. The so-called "holiday
heart" syndrome, where the rate of heart attacks increases immediately following holidays such as New
Years, is a manifestation of this phenomenon.
Research on the effects of alcohol availability on heart disease has lagged behind that on
cirrhosis and traffic accidents. Given the observed relationships between both light and heavy drinking
and taxes on one hand, and between light and heavy drinking and heart disease on the other, we would
expect taxes to play some role in mitigating certain diseases of the circulatory system, and to exacerbate
others.
Drinking may also have some effect on the risk of certain types of cancer. The Physicians’
Health Study, a prospective cohort study of 22 thousand healthy men ages 40 and over, found a Ushaped pattern between all-cause mortality rates and alcohol consumption. The lowest mortality rate
was for light drinkers (2-4 drinks per week) and highest for the group drinking the most (2 or more
drinks per day), with the upturn due to cancer and cardiovascular disease (Camargo et al. 1997)
5.5 Crime and Suicide
Under the influence of alcohol, a parent may be provoked to strike an irritating child; a college
student may forcefully insist on having sex with his date; friends may escalate an argument into a bloody
fight; a robbery victim may foolishly attempt resistance in the face of a loaded gun; soccer fans may riot
in response to an unsatisfactory game. Some individuals under certain circumstances are more prone to
violence, or to provoking violence, when drinking than when sober (Fagan 1990).
33
Drinking affects violent behavior through a number of mechanisms. Drinking may change the
objective consequences of violence, since alcohol acts as an anesthetic and also as an excuse. It may
also act on information-processing capacity, making people myopic and narrowing their repertoire of
responses to a tense situation. It may also cause self-management problems, in which impulse gets the
better of long-term interests (Cook and Moore 1993b).
Economists have not contributed much to the empirical work on this subject. One exception is
Cook and Moore (1993b), in which we examine the effects of aggregate drinking and alcohol taxation
on four forms of violent crime in a state-year panel of data for the years 1979-88. Using a fixed-effects
specification for state and year, we find that per capita alcohol consumption has a significantly positive
effect on rape, aggravated assault, and robbery, and a negligible effect on criminal homicide rates. In
the reduced-form estimates, the state beer-excise tax rate has a strong and significantly negative effect
on rape and robbery, but not on homicide or assault.
There is scant economic research in the areas of domestic violence. Markowitz and Grossman
(1998) examine the effects of beer tax rates, illegal drug prices, and alcohol-control measures on violent
behavior towards children. Their findings include a significant reduction in the likelihood of any violent
behavior to beer-tax increases. There is a weak relationship between the number of licensed retail
alcohol outlets and the likelihood of violence directed at children. More severe acts of violence appear
to be more responsive to tax increases than is violence in general, with elasticities equal to about -.25
across all specifications of the model. Estimated effects of the number of retail outlets likewise indicate
a positive effect on severe violence.
The propensity to commit suicide may also be influenced by heavy drinking. The blood of
suicide victims often contains a high percentage of alcohol (Hayward et al. 1992), and receiving
34
treatment for alcoholism or alcohol abuse is a significant risk factor for suicide (Draper 1994). Skog and
Elekes (1993) examined the relationship between alcohol consumption and suicides in Hungary, and
found the two to be highly positively correlated, with a lag of one year in alcohol consumption.
In an interesting report of a natural experiment, Wasserman et al. (1994) examined the
relationship between male suicides and alcohol consumption in the Soviet Union during the period of
Perestroika, 1985-88. The early years of this period were characterized by a very restrictive alcohol
policy. Relative to the last year of the Brezhnev regime (1984), suicides and violent deaths declined
sharply in 1986, falling to 65 percent of their 1984 level. By 1988, violent deaths were 72 percent of
their 1984 level, and suicides 61 percent. In 1990, the last year covered by the data, these rates of
death due to violence and suicide held at 85 and 68 percent. Meanwhile, total male mortality had
returned to its 1984 level by 1990.
Simple regression analyses of the Soviet data indicate that the suicide rate for men falls by 1.3
per 100,000 for every 1 liter-per-year reduction in pure alcohol consumption per capita. Violent death
rates are more sensitive to alcohol consumption with estimated effects ranging across provinces from 7
to 20 deaths per 100,000, per liter of alcohol.
6. Productivity
The belief that drinking impairs productivity has helped motivate a wide range of both private
and public responses, from workplace rules banning drinking on the job to alcohol regulations governing
the armed forces. National estimates of alcohol-related social costs are typically dominated by the
value of lost productivity. Historically this concern with the quality and quantity of work provided by the
35
labor force was a major factor in Nineteenth Century temperance movements in the United States and
Europe (Roberts 1983; Rumbarger 1989). Clark Warburton (1932) stated the argument concisely:
Prohibition, if it actually resulted in the cessation of use of alcoholic beverages, might be
expected to affect the efficiency of industry in several ways. The principal effect of alcohol is on
the central nervous system, and experiments show that a decrease in the consumption of alcohol
during, or immediately preceding, working hours is accompanied by greater skill at work. The
finer co-ordination made possible by the absence of alcohol tends to reduce the accident rate.
The elimination of drinking bouts should tend to eliminate absenteeism, especially on Monday,
and irregularity in reporting at work. The impossibility of drinking to excess should result in less
sickness and absence on account thereof, and in a longer average working life (195-196).
In recent times the belief that alcohol abuse reduces the productivity of some employees has persuaded
the majority of large corporations in the United States to establish occupational alcoholism programs or
employee-assistance programs (Walsh 1982). Productivity concerns were also paramount in the Soviet
Union in 1985, when Mikhail Gorbachev decreed stringent new controls on production, availability, and
price of alcoholic beverages, which reduced consumption by about one-third in the next couple of
years. There is evidence that on-the-job drinking fell, which may have accounted in part for the rapid
economic growth of that period (The Economist 12/23/89, p. 50).
Curiously, however, the belief that heavy drinking impairs productivity does not receive
unambiguous support from the econometric work on this subject. We begin our review with the
literature focusing on the direct effects of current drinking (or alcohol-related health status) on earnings
or some other indicator of productivity.
36
6.1 Direct effects of drinking on productivity
An early American effort to estimate the productivity costs of drinking is due to Irving Fisher
(1926). His view was that drinking slowed down the "human machine" (p. 118), and he noted that "All
of us know that industrial efficiency was one of the chief reasons for Prohibition (p. 158)." He
supported his claim of impaired productivity by citing a number of experiments, which showed that
drinking reduces proficiency or speed at some task. In particular, he noted an experiment in which four
typesetters were studied over a four-day period; two of them were given drinks, and the other two
were used as a control group. The conclusion was that drinking three glasses of beer in a day reduced
productivity by about 10 percent. Fisher made a heroic extrapolation from this result, projecting a 5
percent increase in national productivity as a result of reduced drinking caused by Prohibition.
Modern scholars studying productivity effects have enjoyed larger sample sizes but unlike Fisher
have utilized nonexperimental data. The typical econometric study estimates the productivity effects of
drinking, utilizing survey data in which respondents are asked about their drinking, work, income, and
other items. The dependent variable is a measure of earnings or hours worked, while the key
independent variable is a measure of the quantity or pattern of contemporaneous drinking, or alcoholrelated psychiatric disorder (alcohol dependence or abuse).
Contemporaneous drinking. The most consistent finding is paradoxical, namely that drinkers
tend to earn more than nondrinkers. For example, based on the Quality of Employment Survey (QES)
data for full-time male American workers in 1972-3, lifetime abstainers and recent abstainers earned
$9,000 and $8,500 respectively; for drinkers, however, earnings were from $10-11,000 across the
drinking spectrum from 1 drink per month to 120 or more (Cook 1991). The pattern that "abstainers
37
earn less" holds up when other characteristics of the individual are controlled for in an OLS regression,
and appears to be true for women as well as men (Berger and Leigh 1988; Bryant et al. 1992; Zarkin et
al. 1998). Some studies find an inverted U-shape between earnings and drinking (French and Zarkin
1995), but others confirm the QES finding that there is a "drinking bonus" at all levels of self-reported
alcohol consumption.
Kenkel and Ribar (1994) provide one of the most thorough explorations of the relationship
between drinking and earnings, although they do not analyze the abstainers as the special case, which
they apparently are. The authors' data are from the National Longitudinal Survey of Youth (NLSY) for
1989 (when the respondents were 24-32 years old). Among their measures of contemporaneous
drinking are the number of days in the past month in which the respondent drank ("days drinking") and
the number of days in which he or she consumed 6 or more drinks ("heavy drinking"). The effects of
"days drinking" on log earnings and log of hours worked was negligible for men and small but discernibly
positive for women, even when a long list of control variables were included. The results for "heavy
drinking" indicate little or no effect for women, and a small negative effect for men.
The authors note three possible sources of bias in OLS results of this sort: omitted variables that
may influence both drinking and earnings; errors in self-reported drinking; and reverse causation,
whereby earnings influence alcohol consumption. They address the latter two problems through a
simultaneous-equations analysis in which the identifying variables are indicators of alcohol availability in
the respondent's state and of alcoholism in his or her family. This specification yields evidence that
"heavy drinking" and "days drinking" reduce male earnings, while increasing female hours and having no
discernible effect on earnings.
Finally, several studies have analyzed the effect of drinking on absenteeism, also with mixed
38
results. Manning et al. (1991) report results from two data sets, the Rand Health Insurance Experiment
(HIE) and the National Health Interview Study for 1983. In neither do they find a relationship between
quantity consumed by current drinkers and absenteeism. (Using the HIE they find that "former drinkers"
have 38 percent higher absentee rates than others.) On the other hand, French and Zarkin (1995),
using survey data for workers at four large work sites, find that both overall drinking and frequency of
drunkenness are positively related to absenteeism.
Alcohol dependence and abuse. Figure 1 above suggests that in addition to the effect of
current drinking on productivity, there may also be an effect of past drinking as mediated by health
status. Several studies have explored this linkage using data on two alcohol-related conditions termed
"alcohol dependence" and "alcohol abuse" in the Diagnostic and Statistical Manual of Mental
Disorders of the American Psychiatric Association. "Alcohol dependence" is defined by symptoms
indicating psychological and physical dependence on alcohol and impaired functioning at work or home.
"Alcohol abuse" is signaled by problems and risky behavior associated with bouts of heavy drinking.
Mullahy and Sindelar (1993) utilize data from the Epidemiological Catchment Area survey of
the New Haven area, which includes diagnostic questions on mental illness. Twenty percent of their
primary sample (males age 30-59) had at some point in their lives experienced the symptoms defining
alcohol dependence or abuse. This group, which the authors term "alcoholics," had a substantially lower
employment rate than the non-alcoholics, and lower earnings -- results that held up after controlling for
other characteristics. Strangely, alcoholic males in their 20s and 60s actually had higher earnings than
nonalcoholics.
Kenkel and Ribar (1994) in their analysis of the NLSY data discussed above, estimate the
effect of current alcohol dependence and abuse. In their OLS estimates they find no evidence that
39
either of these conditions affect hours worked, but report that these conditions suppress earnings for
men and enhance them for women. The simultaneous-equations analysis also finds reduced earnings for
men but no effect (except longer hours) for women.
Kenkel and Wang (1998) extend this analysis of NLSY data by comparing job attributes of
alcoholic and non-alcoholic men. They find that male alcoholics are less likely to receive a variety of
fringe benefits, more likely to be injured on the job, and work for smaller firms. Alcoholics are less
likely to be in white-collar occupations, but those who are earn about as much as non-alcoholics. In
blue-collar professions, however, the alcoholics earn about 15 percent less.
Summary and interpretation. The estimated relationship between self-reported drinking and
measures of the quality and quantity of labor supply (hours worked, earnings, absenteeism) is not what
we might expect. Abstainers clearly earn less, even when we control for a wide range of other
individual characteristics. The evidence on whether the labor market penalizes heavy drinking (without
regard to symptoms of dependence) is at best inconsistent. Prime-age males who exhibit symptoms of
alcohol dependence or abuse earn less, but women in this category appear to earn more.
So were Irving Fisher and Mikhail Gorbachev wrong? Was it a mistake for the U.S. military to
crack down on heavy drinking, and for most corporations to establish employee assistance programs?
We doubt it. More likely some combination of systematic errors in self-reported drinking and
specification error account for the anomalous findings.
In the latter regard, we note the likelihood that the workplace social environment has an
important influence on individual drinking, a causal link that has not been captured in the existing
literature.16 For example, a man who has a taste for working with his hands may become a construction
worker, an occupation that places him in a "wetter" work environment than if he instead had chosen to
40
become an accountant. A woman who enters a male-dominated profession such as law may tend to
take on male drinking patterns, which compared to other women would make her a heavy drinker.
Still, the balance of the evidence favors the view that in moderate-drinking populations, such as
prime-age adults in the U.S. the true direct effect of drinking on productivity is likely to be negative but
small. One source of evidence on this matter is the views of the drinkers themselves. A number of
surveys have asked respondents whether their drinking has caused them any problems (Room et al.
1994). For example, in a national survey conducted in the United States in 1984, 2.9 percent of men
stated that their drinking had harmed their employment opportunities, and 0.8 percent reported they had
lost or nearly lost a job as a result of their drinking (Hilton and Clark 1987). The corresponding
percentages for women were about half those of the men.
On the other hand, the direct effect of drinking on productivity is not the whole story. As
suggested by the lower loop in Figure 1, heavy drinking may have an indirect effect on productivity by
interfering with the process of accumulation of human capital. We summarize the evidence on this
matter below.
6.2 Drinking and human capital.
In their analysis of ECA for New Haven, Mullahy and Sindelar (1989, 1991) found that teenage
alcohol dependence led to early termination of schooling that in turn reduced subsequent income. First
onset of alcoholism's symptoms before age 19 (as reported retrospectively by adult respondents) was
associated with an 11 percent reduction in schooling attainment, controlling for several other
characteristics.
Cook and Moore (1993) suggest two possible mechanisms by which drinking and schooling
41
may interact for adolescents:
(1)
Heavy drinking may interfere with learning and classroom performance, thereby reducing the
contribution to human capital of an additional year of schooling and hence the incentive to
continue;
(2)
To the extent that higher education is rationed according to past scholastic performance and
reputation, heavy drinking may have consequences that increase the cost of continuation (p.
414).
Thus, a forward-looking student would make drinking and schooling decisions together; a myopic
student would make them sequentially; but in either case alcoholic-beverage prices and other
determinants of high-school drinking are a determinant of school persistence.
The authors utilize NLSY data for high-school seniors in 1982, which allow inclusion of an
extensive list of covariates. They find that the beer tax and minimum legal purchase age in the
respondent's state have a direct effect on school persistence, as measured either by the number of years
of college or the likelihood that the respondent will eventually graduate from college.
Schooling is not the only dimension of human capital that may be affected by drinking. Kenkel
and Ribar (1994) find that the likelihood of marriage is negatively affected by heavy drinking and
alcoholism symptoms for both men and women, a finding that holds up well across different
specifications.
These results suggest that much of the effect of drinking on productivity may be indirect,
mediated by the accumulation of human capital. If so, controls on youthful drinking become particularly
important in influencing the course of the economy.
42
7. Evaluation of alcohol taxation and other alcohol-control measures
A number of alcohol-control measures have been subjected to systematic evaluation, as
discussed above and elsewhere (e.g., Edwards et al. 1994). Most of this research has been limited to
estimating the effects of such regulations on alcohol consumption and abuse and the consequences
thereof. But some economists have attempted to take the evaluation effort further by assessing the costs
of specific regulations and by placing a monetary value on the estimated benefits. This effort to
determine by technical means whether a policy innovation is beneficial on balance is controversial, given
that it requires placing a monetary value on life and disability, and also requires explicit judgments
concerning how much deference the public should accord revealed individual preferences concerning
drinking and alcohol-related behavior. While economists cannot necessarily claim full wisdom on such
matters, they can help provide structure to the argument.
Probably the most pervasive government intervention in the alcohol market is the imposition of
specific taxes. The economics of alcohol excise taxes is well developed, and will be the main focus
here.
7.1 Background.
Alcohol excise taxes vary widely across time and space. Historically, the very first internal
revenue measure adopted by the U.S. Congress was an excise tax on whiskey (Hu 1950); a subsequent
increase in that tax from 9 to 25 cents per gallon engendered an armed insurrection. The appropriate
level for alcohol excise taxes remains a contentious issue today in the United States at both the federal
43
and state level. Generally the real values of excise-tax rates have trended downward during the postWar period. They are unit taxes, defined in terms of volume rather than value of the product, and
legislated increases have not kept up with inflation. For example, the federal tax on distilled spirits in
1998 (the equivalent of 21 cents per ounce of ethanol) was about four times higher in 1951.
In the European Union alcohol excises have been one of the most difficult-to-resolve issues in
the tax-harmonization effort, since the tax rates differ widely. Nordic countries in particular have long
used high taxes to restrict drinking, rather than simply as a revenue measure, whereas the wine
producing countries tend to have much lower taxes (Kay and Keen 1986).
Alcohol excises and duties have been an important source of public finance in certain times and
places. This has been particularly true in Russia and the old Soviet Union -- in the early 1980s, taxes on
the liquor trade provided about 13 percent of the state budget (The Economist 12/23/89, p. 50) -although in most advanced countries alcohol taxes constitute less than one percent of the total. Still,
there is widespread acceptance that alcohol should be taxed more heavily than other commodities.
Public-finance theory provides a framework for evaluating alcohol-excise taxes. Application of
the standard criteria of economic efficiency and equity requires some account of the externalities and
health effects of alcohol consumption. Also relevant in practice have been historical comparisons and
comparisons with tax rates in other jurisdictions (Cook 1988).
7.2 Equity Criteria
"Equity" or "fairness" is a standard concern of normative tax theory (Stiglitz 1988; Rosen 1988).
Several general principles are widely accepted
*Equals should be treated equally (the "horizontal equity" criterion);
44
*Households with greater ability to pay should be taxed more heavily than households with less
(the "vertical equity" criterion).
*Households that receive greater benefit from government activities should be taxed more
heavily than households that benefit less (the "benefit" criterion).
Incidence. Implementing any of these principles requires at a minimum some knowledge of the
incidence of alcohol-excise taxes. These taxes are paid by wholesale dealers. The evidence from the
United States suggests that taxes are ultimately passed on to consumers with a markup in the form of
higher retail prices (Center for Science in the Public Interest (CPSI) 1989). Given the imperfect
competition that tends to characterize the relevant markets, it is reasonable to suppose that markups
vary widely, depending on the circumstances of the industry when the tax increase is imposed.
Given that consumers end up paying 100 percent or more of a tax increase, it is necessary in
judging vertical equity that we know how alcohol sales are distributed across income levels.17 In the
United States expenditures on alcoholic beverages increase roughly in proportion to income
(Sammartino 1990), but the volume of alcohol consumed increases little if at all with income. Hence
excise taxes, which are imposed by volume rather than value, are highly regressive.18
This conclusion must be qualified, however. First, the incidence of a tax increase is not
determined solely by who buys the product, but also by how the tax increase affects producers and
sellers (Rosen 1988, p. 266). More generally, an increase in the excise tax rate will tend to have effects
on other markets, which should be taken into account in calculating the incidence. For example, a tax
increase on beer will reduce the traffic accident rate, which in turn will reduce the cost of driving, both
directly and indirectly through reduced insurance rates.
Horizontal equity and user fees. With respect to horizontal equity, the fundamental issue is
45
whether otherwise-similar households should be taxed differently because some purchase more alcohol
than others. The household incidence of alcohol excises is highly concentrated; by one estimate, 6.5
percent of U.S. adults consume half the total alcohol sold (Malin et al. 1982). The equity justification
for imposing the tax on drinkers is as a sort of "user fee," which charges them for the negative
externalities of their drinking.
The characterization of an alcohol excise tax as a user fee is also related to the benefit principle
of tax fairness. People who abuse alcohol benefit from certain government programs more than
nonabusers. For example, lifetime heavy drinkers have elevated morbidity and disability, and hence
claim a disproportionate share of government expenditures on medical care and disability payments
through the Social Security system (Rice et al. 1990). Government revenues from alcohol taxes help
defray these and other alcohol-related public expenses.
Given that the bulk of alcohol-related costs are associated with rare events (most notably traffic
accidents), this "user fee" is akin to an insurance premium. Alcohol taxes differ among individuals in
direct relation to how much they drink, which is a strong predictor of the likelihood of an alcohol-related
problem (Moore and Gerstein 1981, p. 45; Edwards et al. 1994). However excises do not
discriminate with respect to other correlates of alcohol-related problems, such as age, sex, prior history
of drinking and alcohol-related problems, or drinking pattern. A 21-year-old man who drinks 7 beers a
week in a single session and then attempts to drive home pays the same tax as a 40-year-old woman
who drinks one beer with dinner each night. From the actuarial viewpoint, then, this tax is imperfect.
A fairly comprehensive study of the external costs of heavy drinking (Manning et al. 1989,
1991) found that most of these costs are not financed by government in the U.S., but rather by private
insurance companies or by innocent victims. The authors concluded that the present value (using a 546
percent discount rate) of external costs per ounce of ethanol consumed was about 48 cents, double the
average state and federal tax per ounce, which suggests that the current "user fee" is not high enough.19
However, this judgment is based on a more inclusive standard than the benefit criterion, which is limited
to benefits provided by government. The implicit notion is that the drinker should pay regardless of
whether the cost is to a government program or to a private organization or individual.
The analysis by Manning et al. attempted to sort out alcohol-related costs between those that
are borne by the drinker or his or her household (internal) and those that are borne by those outside the
household (external). By this reckoning, earnings lost as a result of heavy drinking are internal costs;
reductions in payroll taxes or claims on Social Security benefits are external. Medical and disability
costs that are reimbursed by insurance are internal if the insurance premium is paid by the drinker and
reflects his true risk status, and external otherwise. Motor-vehicle injuries are also divided between
internal (where the injury is to the drunk driver himself) and external (where someone else is injured in
an accident involving a drunk driver).
The question of where the line should be drawn between internal and external is especially
difficult with respect to intrafamily effects. Family members have individual interests that are sometimes
in conflict, though these differences remain "internal" to the extent that they can be negotiated within the
family (Heien and Pittman 1993; Heien 1995) . Nonetheless there is a clear public stake in preventing
alcohol-induced family violence, child abuse and neglect, and fetal damage. The "fairness" problem with
including these costs in the justification for higher excise taxes is that the other family members may end
up paying twice if the drinker does not change his or her behavior -- the higher tax reduces money left
over for other members of the household.20
Another gray area between internal and external costs is with respect to injuries to passengers
47
of vehicles driven by drunk drivers. If the passengers are consenting adults, then their choice reveals an
ex ante judgment that accepting the ride is preferable to the next-best alternative. This argument
appears to generalize to all users of the road, who presumably know there is some chance of
encountering a drunk driver when they choose to drive on it, and choose to accept that risk. But
voluntary acceptance in this case does not imply that there is no externality. If other users of the road
were able to negotiate directly with the drunk, they would likely find room for a Pareto-improving
bargain that kept the drunk from driving. That is less clear for the case of vehicle passengers, who
presumably do have the opportunity to negotiate with the drunk.
Finally, we note the interesting ambiguity concerning the "drinker should pay" benefit criterion.
Suppose that the alcohol excise tax was increased so that total collections were equal to the external
costs associated with drinking. The increase in these tax rates would cause a reduction in tax collections
from other sources. A standard assumption in revenue estimates is that gross national product is fixed,
and that a tax increase causes a reduction in factor incomes, which in turn will reduce income and
payroll tax collections. Boyd and Seldon (1991), using a computable general equilibrium model,
estimate that an increase in alcohol and tobacco taxes will increase net government revenue by only 60
percent of the increase in collections on those taxes. The "loss" of 40 percent is the result of the tax
consequences of the reallocation of economic activity induced by the tax increases.
48
7.3 Economic Efficiency
If there are negative externalities in the consumption of alcoholic beverages, then in the absence
of government action prices will be "too low" in the sense that at the margin the value of drinks to
consumers will be less than their cost. Taxes on alcohol can then be justified as a mechanism for
internalizing the external costs of alcohol abuse; ideally, the tax on a drink should equal the expected
value of the external cost of consuming that drink.
There is a fundamental difference between this Pigovian principle, which is based on an
efficiency argument, and the equity principle that the "drinker should pay." The total revenue generated
by a Pigovian tax may well be greater than the total external cost, because it reflects the external cost of
the marginal drink rather than of the average (inframarginal) drink.
The efficiency principle, unlike the fairness principle, requires that corrective taxation alter
consumers' behavior. If an increase in the tax (and price) of alcoholic beverages had no effect on the
prevalence of alcohol abuse and its external consequences, then it would not improve economic
efficiency. The evidence presented in section 5 above that excise tax increases reduce motor-vehicle
fatalities is particularly germane.
The application of the corrective tax principle to drinking must account for the fact that the
external cost of a drink differs depending on the personality of the drinker, the time and place of
drinking, how many drinks have been consumed already, and the type of alcoholic beverage.21 It is
possible to institute some crude differentiation in tax rates (e.g., on premise vs. off premise), but the
problem remains. Diamond (1974) demonstrated that when social costs differ with circumstances, then
the value of that uniform tax that maximizes social welfare (under certain assumptions) is equal to a
weighted average of the marginal external costs. Pogue and Sgontz (1989) applied this theory to the
49
case of alcoholic beverages, with a model that assumes all drinkers can be classified as either "abusers"
or "nonabusers." In that model the correct tax depends on the proportion of the drinking population
who are abusers, and their price elasticity of demand relative to nonabusers. We suggest an alternative
approach that does not require this artificial dichotomy and takes better advantage of the empirical
literature: the literature provides estimates of the effects of a change in tax on alcohol consumption, and
on highway fatalities and other damages. The former is the basis for estimating incremental loss of
consumers surplus, while the latter can be used to estimate the reduction in social cost. Taxes should be
increased so long as the latter exceeds the former.
A number of authors have suggested that the principle of corrective taxation be extended to
account for the presumed fact that consumers tend to underestimate the internal costs of their drinking.
If consumers tend to disregard certain costly consequences of their drinking out of ignorance or myopia,
then it is possible that a tax increase would bring their drinking closer into line with their "true"
preferences (Atkinson and Meade 1974; Godfrey and Harrison 1990). The consequence is to increase
the magnitude of the proper corrective tax (Pogue and Sgontz 1989; Phelps 1988).
This analysis of efficiency presumes that the price system functions well in other respects. That
is not the case. For example, taxes on income have the effect of reducing the incentive to provide labor
services to the market. Taxes reduce take-home pay and may distort such choices as how much to
work, how much effort to expend while working, and how much to invest in education and training
(Rosen 1980). One approach to correcting for the disincentive effects of income taxation is to impose
special taxes on commodities that are complements to leisure and substitutes for investment in human
capital (Kay and Keen 1986, p. 88; Slemrod 1990, p. 159; Corlett and Hague 1953). Alcohol is
believed to be one such commodity (although note the mixed evidence reported in section 6).
50
Where the bulk of the external cost results from drunk driving, alcohol-induced violence, and
other behavior that is subject to sanctions, then it might be more efficient to increase the "price" imposed
on violators than to increase the alcohol tax (Kenkel 1996). For the case of drunk driving, that “price”
may include legal and private penalties if apprehended and convicted, as well as the expected cost of
causing an accident, which in the United States may include an increase in insurance-premium rates and
civil liability (Sloan and Githens 1994; Sloan , Reilly Schenzler 1995). Still, imposing stiffer sanctions is
socially costly in itself, and constrained by various practical and ethical considerations.22 And no matter
how stiff the sanctions, there will be some who will ignore them and drive drunk or fight anyway. Given
these concerns, there is a role for an ex ante tax to preempt some of the drinking that would otherwise
lead to trouble.
7.4 Total Social Cost
Closely related to the problem of computing the appropriate excise-tax rate is a problem that
has been of somewhat less interest to economists, calculating the total cost of alcohol abuse. Estimates
of the monetized social burden of a disease or health-related activity have become commonplace in the
public-health literature because of their importance in the policy process. Such estimates are a
precursor to evaluation, since the cost is an indication of the benefit should a comprehensive “cure” be
found. A costly disease, one that causes a large reduction in the overall standard of living, thus appears
to have a strong claim on the public fisc for research and prevention activities. But economists have
served more as critics than as producers of these estimates.
The most prominent estimates for alcohol abuse and other diseases follow the template
developed by a task force of the U.S. Public Health Service (Hodgson and Meiners 1978, 1982). This
51
“cost-of-illness” (COI) method is somewhat at odds with the economists’ normative accounting
framework, primarily because the COI accounting framework is structured around production rather
than consumption.
Several estimates of the social costs of alcohol abuse have been widely disseminated in the
United States (Berry and Boland 1977; Harwood et al. 1984; Rice et al. 1990; Harwood, Fountain,
and Livermore 1998). The most recent of these estimates was $148 billion (in 1993), including $99
billion for lost earnings of those who died prematurely or were disabled due to alcohol-related accidents
or disease, and $19 billion for medical care and rehabilitation. As is traditional with this approach, no
account is taken of the subjective value that individuals place on their life and health, and on the lives of
those they care about. Some authors have modified the COI approach to incorporate a willingness-topay-based valuation for additional life years; Miller et al. (1998), for example, estimate the
“comprehensive” social cost of alcohol-involved crashes in the United States as $115 billion in 1993,
much of which stems from the subjective value of lost life and ability.
Even with this addition of the “consumption” value of life, the COI framework does not fit the
standard economic framework because it fails to distinguish between internal and external costs. As we
saw in the discussion of a cost-based excise tax, this distinction is highly relevant, since otherwise the
drinker would be asked to pay the internal costs twice. However, in other applications it may be
appropriate to compute the total social cost, rather than only the external cost. For example, the value
of a cure for liver cirrhosis should include the cirrhosis-related costs currently borne by drinkers (and
their households), as well as those borne by the rest of society.
Finally, estimates of the cost of alcohol abuse are also subject to what might be called
“conceptual” uncertainty, a lack of clarity about the conditions under which the estimated “cost” would
52
be “saved.” For example, estimates of productivity losses due to excess drinking typically presume that
the socioeconomic characteristics of the labor force are not influenced by alcohol abuse. The implicit
question answered by the estimated productivity loss is this: How much would the quality and quantity
of labor increase if some (magical) intervention were instituted that eliminated alcohol abuse among the
working-age population while having no effect on that population's socioeconomic characteristics? Yet
according to the results reviewed in Section 5 above, the primary mechanism by which alcohol abuse
influences productivity is through workers’ marital status and formal education. The general problem
here is that alcohol plays a diverse and complex role in shaping everything from personal health-related
habits and occupational choice to family life and social intercourse. It seems likely that any intervention
that reduced or eliminated alcohol abuse would have a number of ramifications for other aspects of life:
What, then should be held constant in estimating the costs of alcohol abuse?
A number of economists (Myrdal 1930; _sterberg 1983) have suggested that it would make
more sense to estimate the costs and benefits of specific real-world interventions (such as a marginal
increase in the excise tax rate), rather than the imaginary, perfectly effective intervention that underlies
the estimate of total costs. Knowing the intervention helps guide the evaluators in deciding which causal
mechanisms to explore and which to ignore. While this perspective seems valid, there nonetheless
remains a considerable demand for estimates of the total costs of alcohol abuse for use in influencing the
political and policy agenda.
8. Concluding Observations
Economists' contributions to the study of the health consequences of drinking have included
53
both theoretical and empirical advances, and have spanned both the positive and normative realms.
Arguably the most important contribution has been the repeated empirical demonstration that consumers
as a whole are responsive to the general price level of alcoholic beverages. The demand curves for
beer, wine and spirits are negatively sloped; more surprisingly the consumption of ethanol is negatively
related to the prices of the beverages that contain it, despite the obvious opportunity for substitution
among beverage types. As a result, an increase in the price level of alcoholic beverages tends to reduce
the prevalence of alcohol abuse and its health-related consequences. Thus the excise tax on alcoholic
beverages is an effective alcohol-control measure that can be used to promote the public health.
Economists have been somewhat at odds with the public-health community in arguing that the
effectiveness of a government alcohol-control measure is not sufficient justification for imposing it. In
this area, as in others, consumer sovereignty should be given its due, and in setting policy the social
costs of drinking should be balanced against the pleasure to consumers. As we have seen, the
normative literature has included a debate over how fully to respect the revealed preferences of
consumers and households, given that consumers are not necessarily well informed about the
consequences of drinking, that ethanol is addictive, and that drinkers may take inadequate account of
their dependents' welfare.
Further economic research is needed to sort out the costs and benefits of drinking. As we write
this (in 1998), the empirical literature on productivity effects is particularly unsatisfactory; how can we
explain the persistent findings, at least for U.S. data, that drinkers earn more than abstainers, or that for
women who drink (but not men) earnings increase with the amount of alcohol consumed? Economists
have for the most part simply ignored another important issue, the apparent health benefits of moderate
drinking.
54
Outside of the policy arena, alcohol remains interesting for economists because it provides a
clear example of a commodity whose consumption has important intertemporal effects, and which is
subject to social influence. The theory of rational addiction has provided one framework for exploring
the first mechanism; there is also an opportunity here to collaborate with psychologists in an effort to
reconcile the assumptions of the model with experimental evidence on intertemporal decisionmaking.
Meanwhile, economists are just beginning the study of social influence in drinking, another area which
may reward cross-disciplinary collaboration.
The primary justification for public support of research on drinking and its consequences is the
importance to public health worldwide. Economists have played a relatively small but important role in
this research program, challenging some established beliefs about the singularity of alcohol and offering a
broader normative framework. We believe that a closer collaboration between economists and other
behavioral scientists will pay off both in terms of scientific progress and policy influence.
55
References
Adams Business Media (1997), The Beer Handbook (Adams Business Media, Cleveland, OH).
Atkinson, A.B. and T.W. Meade (1974), "Methods and preliminary findings in assessing the economic
and health consequences of smoking with particular reference to lung cancer",
Journal of the Royal Statistical Society Series A 137:297-312.
Babor, T.F. (1985), "Alcohol, economics and the ecological fallacy: Toward an integration of
experimental and quasi-experimental research", in: E. Single and T. Storm, eds., Public Drinking and
Public Policy (Addiction Research Foundation, Toronto) 161-189.
Babor, T.F., J.H. Mendelson, I. Greenberg, and J. Kuehnle (1978), "Experimental analysis of the
'happy hour': Effects of purchase price on alcohol consumption", Psychopharmacology 58:35-41.
Baumann K.E. and S.T. Ennett (1996), "On the importance of peer influence for adolescent drug use:
commonly neglected considerations", Addiction 91(2):185-198.
Becker, G.S. (1992), "Habits, addictions, and traditions", Kyklos 45:327-346.
Becker, G. S. (1996), Accounting for Tastes (Harvard University Press, Cambridge).
Becker, G.S., M. Grossman and K.M. Murphy (1991), "Rational addiction and the effect of price on
consumption", American Economic Review 81(2):237-241.
Becker, G.S. and K.M. Murphy (1988), "A theory of rational addiction", Journal of Political Economy
96(4):675-700.
Ben-Porath, Y. (1967), "The production of human capital and the life cycle of earnings", Journal of
Political Economy 75:352-365.
Berger, M.C. and J.P. Leigh (1988), "The effect of alcohol use on wages", Applied Economics
20:1343-1351.
Berry, R.E. and J.P. Boland (1977), The Economic Cost of Alcohol Abuse (The Free Press, New
York).
Boland, B. and R. Roizen (1973), "Sales slips and survey responses: New data on the reliability of
survey consumption measures", Drinking and Drug Practices Surveyor 8:5-10.
Boyd, R. and B.J. Seldon (1991), "Changes in sin taxes: A general equilibrium perspective", Land
Economics 67(3):365-374.
Bray, R.M., M.E. Marsden, J.R. Harbold and M.R. Peterson (1992), "Progress toward eliminating
drug and alcohol abuse among U.S. military personnel", Armed Forces & Society 18(4):476-496.
56
Bruun, K., G. Edwards, M. Lumio, K. Makela, L. Pan, R.E. Popham, R. Room, W. Schmidt, O.
Skog, P. Sulkunen and E. _sterberg (1975), Alcohol Control Policies in Public Health Perspective
(Aurasen Kirjapaino, Forssa, Finland).
Bryant, R.R., V.A. Sumaranayake and A. Wilhite (1992), "Alcohol use and wages of young men:
Whites vs. nonwhites", International Review of Applied Economics 6(2):184-202.
Camargo, C.A. Jr, C.H. Hennekens, J.M. Gaziano, R.J. Glynn, J.E. Manson, M.J. Stampfer (1997),
"Prospective study of moderate alcohol consumption and mortality in US male physicians", Archives of
Internal Medicine 157(1):79-85.
Center for Science in the Public Interest (1989), Impact of Alcohol Excise Tax Increases on Federal
Revenues, Alcohol Consumption and Alcohol Problems (Center for Science in the Public Interest,
Washington, DC).
Centers for Disease Control (1990), "Alcohol-Related Mortality and Years of Potential Life Lost United States, 1987", Morbidity and Mortality Weekly Report 39(11), March 23.
Chaloupka, F.J. and M. Grossman (1994), "Drinking and addiction? Evidence from the first National
Health and Nutrition Examination Survey" University of Illinois Chicago.
Chaloupka, F.J. and H. Wechsler (1996), "Binge drinking in college: The impact of price",
Contemporary Economic Policy 14(4):112-124.
Chaloupka, F.J. and A. Laixuthai (1997), "Do youths substitute alcohol and marijuana? Some
econometric evidence", Eastern Economic Journal 23(3):253-276.
Chaloupka, F.J., H. Saffer and M. Grossman (1993), "Alcohol control policies and motor vehicle
fatalities", Journal of Legal Studies 22(1):161-186.
Chaloupka, F.J., M. Grossman, G.S. Becker and K.M. Murphy (1993), "Alcohol addiction: An
econometric analysis", Presented at the annual meeting of the American Economic Association,
Anaheim, CA, January.
Chamberlain, G. (1984), "Panel data" in Z. Grilliches and M. Intrilligator, eds., Handbook of
Econometrics (North Holland, New York).
Chesson, H.W., P. Harrison and W.J. Kassler (1997), "Alcohol, youth, and risky sex: The effect of
beer taxes and the drinking age on gonorrhea rates in teenagers and young adults", Working Paper.
Clements, K.W., W. Yang, and S.W. Zheng (1997), "Is utility additive? The case of alcohol Applied
Economics 29: 1163-1167.
57
Coate, D. and M. Grossman (1988), "Effects of alcoholic beverage prices and legal drinking ages on
youth alcohol use", Journal of Law and Economics 31(1):145-171.
Collins, D. and H. Lapsley (1996), The social costs of drug abuse in Australia in 1988 and 1992.
(Commonwealth Department of Human Services and Health, Canberra, Australia).
Collins, J.J. Jr. (1989), "Alcohol and interpersonal violence: Less than meets the eye", in: N.A. Weiner
and M.E. Wolfgang, eds., Pathways to Criminal Violence (Sage Publications, Newbury Park, CA) 4967.
Cook, P.J. (1981), "The effect of liquor taxes on drinking, cirrhosis, and auto fatalities", in: M. Moore
and D. Gerstein, eds., Alcohol and Public Policy: Beyond the Shadow of Prohibition (National
Academy of Sciences, Washington DC) 255-285.
Cook, P.J. (1988), "Increasing the federal excise taxes on alcoholic beverages" Journal of Health
Economics, March: 89-91.
Cook, P.J. (1991), "The social costs of drinking", in: Aasland, A.G., ed., The Expert Meeting on the
Negative Social Consequences of Alcohol Abuse (Norwegian Ministry of Health and Social Affairs,
Oslo, Norway) 49-81.
Cook, P.J. and J. Leitzel (1996), "Perversity, futility, jeopardy: An economic analysis of the attack on
gun control" Law and Contemporary Problems 59(1): 91-118.
Cook, P.J. and M.J. Moore (1993a), "Taxation of alcoholic beverages", in: M. Hilton and G. Bloss,
eds., Economic Research on the Prevention of Alcohol-Related Problems (National Institute on Alcohol
Abuse and Alcoholism, NIH Publication No. 93-3513, Rockville, MD) 33-58.
Cook, P.J. and M.J. Moore (1993b), "Economic perspectives on alcohol-related violence", in: S.E.
Martin, ed., Alcohol-Related Violence: Interdisciplinary Perspectives and Research Directions (National
Institute on Alcohol Abuse and Alcoholism, NIH Publication No. 93-3496, Rockville, MD) 193-212.
Cook, P.J. and M.J. Moore (1993c), "Drinking and schooling", Journal of Health Economics
12:411-429.
Cook, P.J. and M.J. Moore (1994), "This tax's for you", National Tax Journal September:559-573.
Cook, P.J., M.J. Moore and R. Pacula (1993), Drinking by Young Adults, Part I: Demographics
(Duke University, Durham, North Carolina, November).
Cook, P.J. and G. Tauchen (1982), "The effect of liquor taxes on heavy drinking", Bell Journal of
Economics 13(2):379-390.
Cook, P.J. and G. Tauchen (1984), "The effect of minimum drinking age legislation on youthful auto
fatalities, 1970-77", Journal of Legal Studies 13:169-190.
58
Corlett, W.J., and D.C. Hague (1953), "Complementarity and the excess burden of taxation", Review
of Economic Studies 21:21-30.
Criqui, M.H., L.D. Cowan, H.A. Tyroler, S. Bangdiwala, G. Heiss, R.B. Wallace, and R. Cohn
(1987), "Lipoproteins as mediators for the effects of alcohol consumption and cigarette smoking on
cardiovascular mortality: Results from the Lipid Research Clinics Follow-up Study", American Journal
of Epidemiology 126:629-637.
DeBakey S.F., F.S. Stinson, B.F. Grant, M.C. Dufour (1995), Liver Cirrhosis Mortality in the United
States, 1970-92 (National Institute on Alcohol Abuse and Alcoholism, Washington DC).
Diamond, P. (1974), Consumption externalities and imperfect corrective pricing", Bell Journal of
Economics 5: 526-538.
DiNardo, J. and T. Lemieux (1992), "Alcohol, marijuana, and American youth: The unintended effects
of government regulation", National Bureau of Economic Research Working Paper No. 4212,
November.
Draper, B. (1994),"Suicidal behavior in the elderly", International Journal of Geriatric Psychiatry
9:655-661.
Duan, N. et al. (1984), "Choosing between the sample-selection model and the multi-part model",
Journal of Business and Economic Statistics 2(3):283-289.
Dufour, M.C. (1996) "Risks and benefits of alcohol use over the life span", Alcohol Health and
Research World 20(3):145-151.
DuMouchel, W.A., A.F. Williams and P. Zador (1987), "Raising the alcohol purchase age: Its effect on
fatal motor crashes in 26 states", Journal of Legal Studies 16:249-266.
Edwards, G. et al (1994), Alcohol Policy and the Public Good (Oxford University Press, New York).
Fagan, J. (1990), "Intoxication and aggression", in: M. Tonry and J.Q. Wilson, eds., Crime and Justice,
Vol. 13 (University of Chicago Press, Chicago) 241-320.
Fisher, I. (1926), Prohibition at its Worst (The MacMillan Company, New York).
French, M.T. and G.A. Zarkin (1995), "Is moderate alcohol use related to wages? Evidence from Four
worksites", Journal of Health Economics 14(3):319-344.
Gaviria, A. and S. Raphael (1997), "School-based peer effects and juvenile behavior" University of
California at San Diego, Working Paper.
59
Gerstein, D.R. (1981), "Alcohol use and consequences", in: M.H. Moore and D.R. Gerstein eds.,
Alcohol and Public Policy: Beyond the Shadow of Prohibition (National Academy of Sciences,
Washington DC) 182-224.
Gius, M.P. (1996), "Using panel data to determine the effect of advertising on band-level distilled
spirits sales", Journal of Studies on Alcohol 57(1):73-76.
Godfrey, C. and L. Harrison (1990), "Preventive health objectives and tax policy options", in: A.
Maynard and P. Tether eds. Preventing Alcohol and Tobacco Problems. Vol.1. (Avebury, Gower
Publishing Co., Hants, England) 54-74.
Grossman, M. (1972), "On the concept of health capital and the demand for health," Journal of Political
Economy 80:223-255.
Grossman, M. (1993), "The economic analysis of addictive behavior", in: M. Hilton and G. Bloss, eds.,
Economics and the Prevention of Alcohol-Related Problems (National Institutes of Health, Washington,
DC) NIH Publication No. 93-3513, 91-124.
Grossman, M., F.J. Chaloupka, H. Saffer and A. Laixuthai (1994), "Effects of alcohol price policy on
youth: A summary of economic research", Journal of Research on Adolescence 4:347-364.
Grossman, M., F.J. Chaloupka and I. Sirtalan (1998), "An empirical analysis of alcohol addiction:
Results from the monitoring the future panels," Economic Inquiry 36:39-48.
Grossman, M., D. Coate and G.M. Arluck (1987), "Price sensitivity of alcoholic beverages in the
United States: Youth alcohol consumption", in: Holder, H., ed., Control Issues in Alcohol Abuse
Prevention: Strategies for States and Communities (JAI Press, Inc., Greenwich) 169-198.
Grossman, M., F. Chaloupka, G. Becker, and K. Murphy (1992), "Alcohol addiction: An econometric
analysis", Working Paper.
Grunewald, P.J., A.B. Millar, and P. Roeper (1996), "Access to alcohol: Geography and Prevention for
Local Communities", Alcohol Health & Research World 20(4):244-251.
Harford, T.C. and S.D. Brooks (1992), "Cirrhosis mortality and occupation", Journal of Studies on
Alcohol 53:463-468.
Harwood, H., D. Fountain, and G. Livermore (1998), The Economic Costs of Alcohol and Drug
Abuse in the United States 1992. (U.S. Government Printing Office, Washington, DC).
Harwood, H.J., D.M. Napolitano, P.L. Kristiansen and J.J. Collins, eds. (1984), Economic Costs to
Society of Alcohol and Drug Abuse and Mental Illness: 1980 (Research Triangle Institute, Research
Triangle Park, NC).
Hauge, R. (1988), "The effects of changes in availability of alcoholic beverages", in: M.D. Laurence,
60
J.R. Snortum and F.E. Zimring, eds., Social Control of the Drinking Driver (University of Chicago
Press, Chicago) 169-188.
Hayward, L., S.R. Zubrick, and S. Silburn (1992),"Blood alcohol levels in suicide cases", Journal of
Epidemiology and Community Health 46:256-60.
Heckman, J.J. (1981), "Heterogeneity and state dependence", in: S. Rosen, ed., Studies in Labor
Markets (University of Chicago Press, Chicago).
Heien, D.M. and D.J. Pittman (1993), "The external costs of alcohol abuse", Journal of Studies on
Alcohol 54:302-307.
Heien, D.M. (1995), "The economic case against higher alcohol taxes", Journal of Economic
Perspectives, 9(1) Winter:207-209.
Hilton, M.E. and W.B. Clark (1987), "Changes in American drinking patterns and problems,
1967-1984", Journal of Studies on Alcohol 48(6):515-522.
Hodgson, T.A. and M. Meiners (1979), Guidelines for Cost-of-Illness Studies in the Public Health
Service (Public Health Service Task Force on Cost-of-Illness Studies, Bethesda, MD).
Hodgson, T.A. and M. Meiners (1982), "Cost-of illness methodology: A guide to current practices and
procedures", Milbank Memorial Fund Quarterly 60(3):429-462.
Hoyt, G.M. and F.J. Chaloupka (1994), "Effect of survey conditions of self-reported substance use",
Contemporary Economic Policy 12(3):109-121.
Hu, T-Y (1950), The Liquor Tax in the United States, 1791-1947. (Columbia University Press, New
York).
Kay, J. and M. Keen (1986), "Alcohol and tobacco taxes: Criteria for harmonisation", in: S Cnossen,
ed., Tax Coordination in the European Community (Kluwer Academic, Amsterdam) 85-111.
Keller, M. (1978), "Alcohol consumption", in: The New Encyclopaedia Britannica in 30 Volumes:
Macropaedia Vol. 1, 15th ed. (Encyclopedia Britannica, Inc., Chicago) 437-450.
Kenkel, D.S. (1993), "Drinking, driving, and deterrence: The effectiveness and social costs of
alternative policies", Journal of Law and Economics 36:877-913.
Kenkel, D.S. (1996), "New estimates of the optimal tax on alcohol", Economic Inquiry 34:296-319.
Kenkel, D.S. (1998), "Price and tax effects on demand and health outcomes", in: National Institute on
Alcohol Abuse and Alcoholism Tenth Special Report to the US Congress on Alcohol and Health, in
press.
61
Kenkel, D.S. and D. Ribar (1994), "Alcohol consumption and young adults' socioeconomic status", in:
Brookings Papers on Economic Activity-Micro June:119-161.
Kenkel, D.S. and P. Wang (1998), "Are alcoholics in bad jobs?", National Bureau of Economic
Research Working Paper 6401.
Klatsky, A.L., M.A. Armstrong, and G.D. Friedman (1990),"Risk of cardiovascular mortality in
alcohol drinkers, ex-drinkers and nondrinkers", American Journal of Cardiology 66(17):1237-1242.
Lands, W.E.M. and S. Zakhari (1991), "Alcohol and cardiovascular disease", Alcohol Health &
Research World 14(4):304-312.
Larkby C. and N. Day (1997), "The effects of prenatal alcohol exposure", Alcohol Health & Research
World 21(3):192-198.
Lelbach, K. (1974), "Organic pathology related to volume and pattern of alcohol use", in: Gibbins,
R.J., et al., eds., Research Advances in Alcohol and Drug Problems, Vol. 1. (Wiley and Sons, New
York).
Leon, D.A., L. Chenet, V.M. Shkolnikov, S. Zakharov, J. Shapiro, G. Rakhmanova, S. Vassin, M.
McKee (1997), "Huge variation in Russian mortality rates 1984-94: Artifact, alcohol, or what?", The
Lancet 350: 383-388.
Leung, S and C.E. Phelps (1993), "My kingdom for a drink...? A review of estimates of the price
sensitivity of demand for alcoholic beverages", in M.E. Hilton and G. Bloss eds., Economics and the
Prevention of Alcohol-Related Problems. NIAAA Research Monograph No. 25. NIH Pub. No. 933513. (National Institutes of Health, Bethesda, MD) 1-31.
Lyon, A.B. and R.M. Schwab (1995), "Consumption taxes in a life-cycle framework: Are sin taxes
regressive?", Review of Economics and Statistics 77(3):389-406.
Males, M.A. (1986), "The minimum purchase age for alcohol and young-driver fatal crashes: A
long-term view", The Journal of Legal Studies 15(1):181-211.
Malin, H., J. Coakley, C. Kaelber, N. Munch, and W. Holland (1982), "An epidemiologic perspective
on alcohol use and abuse in the United States", in: Alcohol Consumption and Related Problems
(National Institute of Alcohol Abuse & Alcoholism, Washington, DC).
Manning, W., L. Blumberg and L. Moulton (1995), "The demand for alcohol: The differential response
to price", Journal of Health Economics 14(2):123-148.
Manning, W.G., E.B. Keeler, J.P. Newhouse, E.M. Sloss and J. Wasserman (1989), "The taxes of sin:
Do smokers and drinkers pay their way?", Journal of the American Medical Association
62
261:1604-1609.
Manning, W.G., E.B. Keeler, J.P. Newhouse, E.M. Sloss and J. Wasserman, eds. (1991), The Costs
of Poor Health Habits (Harvard University Press, Cambridge).
Manning, W.G. and J. Mullahy (1998), "The effects of price on patterns of alcohol consumption". etc.
Speech presented at the annual meeting of the Allied Social Science Association, Chicago, January
1998.
Manski C.F. (1995), Identification Problems in the Social Sciences (Harvard University Press,
Cambridge).
Markowitz, S. and M. Grossman (1998), "Alcohol regulation and violence towards children", National
Bureau of Economic Research, Working Paper No. 6359.
Mello, N.K. (1972), "Behavioral studies of alcoholism", in: B. Kissin and H. Begleiter, eds., The
Biology of Alcoholism, vol. 2 (Plenum, New York).
Mello, N.K., H.B. McNamee, and J.H. Mendelson (1968), "Drinking patterns of chronic alcoholics:
gambling and motivation for alcohol", in: J.O. Cole, ed., Clinical Research in Alcoholism. (American
Psychiatric Association, Washington, DC).
Midanik, L., (1982), "The validity of self-reported alcohol consumption and alcohol problems: A
literature review", British Journal of Addiction 77:357-382.
Miller, T.R. and L.J. Blincoe (1993), "Incidence and costs of alcohol-involved crashes in the United
States", Accident Analysis & Prevention 26: 583-592.
Miller, T.R., D.C. Lestina and R.S. Spicer (1998), "Highway crash costs in the United States by driver
age, blood alcohol level, victim age, and restraint use", Accident Analysis and Prevention 30(2):137150.
Miller-Tutzauer, C., K.E. Leonard and M. Windle (1991), "Marriage and alcohol use: A longitudinal
study of 'maturing out'", Journal of Studies on Alcohol 52(5):434-440.
Miron, J.A. (1997), "The effect of alcohol prohibition on alcohol consumption", Boston University
working paper.
Miron, J.A. and J. Zwiebel (1991), "Alcohol consumption during prohibition," AEA Papers and
Proceedings 81(2):242-247.
Moore, M.J. (1996), "Death and tobacco taxes", RAND Journal of Economics 27(2):415-428.
Moore, M.J. and P.J. Cook (1995), "Habit and Heterogeneity in the Youthful Demand for Alcohol",
63
National Bureau of Economic Research, Working Paper No. 5152.
Mullahy, J. and J.L. Sindelar (1989), "Life cycle effects of alcoholism on education, earnings, and
occupation", Inquiry 26:272-282.
Mullahy and Sindelar (1991), "Gender differences in labor market effects of alcoholism", American
Economic Review 81(2):161-165.
Mullahy, J. and J.L. Sindelar (1993), "Alcoholism, work, and income", Journal of Labor Economics
11(3):494-520.
Mullahy, J. and J.L. Sindelar (1996), "Employment, unemployment, and problem drinking",
Journal of Health Economics 15(4):409-434.
Mullahy, J. and J.L. Sindelar (1997), "Women and work: tipplers and teetotalers", Health Economics
6(5):533-537.
Myrdal G. (1930), Alkoholens vinst – och Forlustkonto. Tirfling 24(4):106-109.
National Household Survey on Drug Abuse (1996), Online data from Prevline, Prevention Online web
site, from The National Clearinghouse for Alcohol and Drug Information
(www.health.org/pubs/nhsda/96hhs/tabtoc.htm) .
Norström, T. (1987), "The abolition of the Swedish alcohol rationing system: Effects on consumption
distribution and cirrhosis mortality", British Journal of Addiction 82:633-642.
Norton E.C., Lindrooth R.C. and Ennett StT. (In press), "Controlling for the endogeneity of peer
substance use on adoolescent alcohol and tobacco use", Health Economics.
NTC Publications (1996), World Drink Trends: International Beverage Alcohol Consumption and
Production (NTC Publications ,Trends Henley-on-Thames, Oxfordshire, England).
Ohsfeldt, R.L. and M.A. Morrisey (1997), "Beer taxes, workers' compensation, and industrial injury",
Review of Economics and Statistics 79(1):155-160.
_sterberg E. (1983), "Calculating the costs of alcohol: The Scandinavian experience", in: M. Grant, M.
Plant, and A. Williams, eds., Economics and Alcohol: Consumption and Controls. (Gardner Press,
New York) 82-96.
Pacula R.L. (1998), Does increasing the beer tax reduce marijuana consumption?", Journal of Health
Economics 17(5):557-586.
Pernanen, K. (1991), Alcohol in Human Violence (Guilford Press, New York).
Phelps, C.E. (1988), "Death and taxes: an opportunity for substitution", Journal of Health Economic
7:1-24.
64
Pogue, T.F. and L.G. Sgontz (1989), "Taxing to control social costs: The case of alcohol", American
Economic Review 79(1):235-243.
Polich, M.J. and B.R. Orvis (1979), Alcohol Problems: Patterns and Prevalence in the U.S. Air Force
(The Rand Corporation, Santa Monica, CA).
Pollak, R.A. (1970), "Habit formation and dynamic demand functions," Journal of Political Economy
78:745-763.
Popham, R.E. and W. Schmidt (1981), "Words and deeds: The validity of self-report data on alcohol
consumption", Journal of Studies on Alcohol 42(3):355-358.
Pudney, S. (1989), Modeling Individual Choice: The Econometrics of Corners, Kinks, and Holes
(Basil-Blackwell, Oxford).
Ramsey, F.P. (1927), "A contribution to the theory of taxation", Economic Journal 37(1):47-61
.
Rice, D.P., S. Kelman, L.S. Miller and S. Dunmeyer (1990), The Economic Costs of Alcohol and Drug
Abuse and Mental Illness: 1985, Report submitted to the Office of Financing and Coverage Policy of
the Alcohol, Drug Abuse, and Mental Health Administration, (US Department of Health and Human
Services, Institute for Health & Aging, University of California, San Francisco).
Rimm, E.B., E.L. Giovannucci, W.C. Willett, G.A. Colditz, A. Ascherio, B. Rosner, M.J. Stampfer
(1991), "Prospective study of alcohol consumption and risk of coronary disease in men", The Lancet
338:464-68.
Roberts, J.S. (1984), Drink, Temperance and the Working Class in Nineteenth Century Germany
(Allen & Unwin, Boston).
Room, R., S. Bondy and J. Ferris (1994), "Drinking and risk of alcohol-related harm in a 1989
Canadian national sample", Working Paper (Addiction Research Foundation, Toronto, Ontario).
Rosen, H.S. (1980), "What is labor supply and do taxes affect it?", American Economic Review Papers
and Proceedings 70(2):171-176.
Rosen, H.S. (1988), Public Finance, 2nd ed. (Irwin, Homewood, IL).
Ruhm, C.J. (1995), "Economic conditions and alcohol problems", Journal of Health Economics
14(5):583-603.
Ruhm, C.J. (1996), "Alcohol policies and highway vehicle fatalities", Journal of Health Economics
15:435-454.
Rumbarger J.J. (1989), Profits, Power, and Prohibition: Alcohol Reform and the Industrializing of
America 1800-1930 (State University of New York Press, Albany).
65
Saffer, H. (1991), "Alcohol advertising bans and alcohol abuse: An international perspective", Journal of
Health Economics 10(1):65-79.
Saffer, H. (1994), "Alcohol tax equalization and social costs", Eastern Economic Journal, 20(1)
Winter:33-43.
Saffer, H. (1995), "Alcohol advertising and alcohol consumption: Econometric studies", in: Martin, S.E.,
ed. The Effects of the Mass Media on the Use and Abuse of Alcohol. NIAAA Research Monograph
No. 28. NIH Pub. No. 95-3743. (National Institutes of Health, National Institute on Alcohol Abuse
and Alcoholism, Bethesda, MD) 83-99.
Saffer, H. (1997), "Alcohol advertising and motor vehicle fatalities", Review of Economics and Statistics
79(3): 431-442.
Saffer, H. and M. Grossman (1987a), "Beer taxes, the legal drinking age, and youth motor vehicle
fatalities", Journal of Legal Studies 16:351-374.
Saffer, H. and M. Grossman (1987b), "Drinking age laws and highway mortality rates: Cause and
effect", Economic Inquiry 25(3):403-417.
Sammartino, F. (1990), Federal Taxation of Tobacco, Alcoholic Beverages, and Motor Fuels
(Congressional Budget Office, Washington DC).
Shaper, A.G. (1990), "Alcohol and mortality: a review of prospective studies", British Journal of
Addiction 85:837-847.
Shaper, A.G., G. Wannametheee, and M. Walker (1988), "Alcohol and mortality in British men:
Explaining the U-shaped curve ", Lancet 2(8623):1267-1273.
Shavell, S. (1986), "The judgment-proof problem", International Review of Law and Economics 6: 45.
Sindelar, J.L. (1993), "Measurement issues in alcohol survey data", in: M.E. Hilton and G Bloss, eds.,
Economics and the Prevention of Alcohol-Related Problems. NIH Pub. No. 93-3513. (National
Institute on Alcohol Abuse and Alcoholism, Rockville, MD) 201-228.
Skog, O-J. (1980), "Social interaction and the distribution of alcohol consumption", Journal of Drug
Issues Winter:71-92.
Skog, O-J. (1985), "The collectivity of drinking cultures: A theory of the distribution of alcohol
consumption", British Journal of Addiction 80:83-99.
Skog, O-J, and Z. Elekes (1993),"Alcohol and the 1950-90 Hungarian suicide trend-Is there a causal
connection?", Acta Sociologica 36:33-46.
66
Slemrod, J. (1990), "Optimal taxation and optimal tax systems", Journal of Economic Perspectives
4(1):157-178.
Sloan, F.A. and P.B. Githens (1994), "Drinking, driving, and the price of automobile insurance", Journal
of Risk and Insurance 61(1):33-58.
Sloan, F.A., B.A. Reilly and C. Schenzler (1995), "Effects of tort liability and insurance on heavy
drinking and drinking and driving", Journal of Law and Economics 38:49-77.
Stampfer, M.J.,G.A. Colditz, W.C. Willett, F.E. Speizer, and C.H. Hennekens (1988), "Prospective
study of moderate alcohol consumption and the risk of coronary disease and stroke in women", New
England Journal of Medicine 319:267-273.
Stiglitz, J.E. (1988), Economics of the Public Sector. 2nd ed. (W.W. Norton & Co, New York).
Stinson, F.S., J.D. Lane, G.D. Williams and M.C. Dufour (1997), U.S. Apparent Consumption of
Alcoholic Beverages Based on State Sales, Taxation, or Receipt Data in: U.S. Alcohol Epidemiologic
Data Reference Manual (National Institute of Alcohol Abuse & Alcoholism, Division of Biometry and
Epidemiology, Alcohol Epidemiologic Data System, Rockville, MD).
Sulkunen, P. (1983), "Alcohol consumption and the transformation of living conditions: A comparative
study", in: Smart, R.G., F.B. Glaser, Y. Israel, H. Kalant, R.E. Popham and W. Schmidt, eds.,
Research Advances in Alcohol and Drug Problems, Vol. 7 (Plenum, New York) 247-297.
Trice, H.M. and W.J. Sonnenstuhl (1988), "Drinking behavior and risk factors related to the work
place: Implications for research and prevention", The Journal of Applied Behavioral Science
24(4):327-346.
U.S. Department of Health and Human Services (1997), Ninth Special Report to the U.S. Congress on
Alcohol and Health (U.S. Government Printing Office, Washington, DC).
Vallee B.L. (1998), "Alcohol in the western world", Scientific American 278(6):80-85.
Wales, T.J., and A.D. Woodland (1983),"Estimation of consumer demand systems with binding nonnegativity constraints", Journal of Econometrics, 21:437-468.
Walsh, D.C. (1982), "Employee assistance programs", Milbank Memorial Fund Quarterly/Health and
Society, 60(3): 492-517.
Warburton, C. (1932), The Economic Results of Prohibition (Columbia University Press, New York).
Wasserman, D, A. Varnik, and G. Eklund (1994),"Male suicides and alcohol consumption in the
former USSR", Acta Psychiatrica Scandinavica 89:306-313.
Williams, G.D., F.S. Stinson, J.D. Lane, S.L. Tunson, M.C. Dufour (1996), Surveillance report #39:
67
Apparent per capita alcohol consumption: National, state, and regional trends, 1977-94. (National
Institute on Alcohol Abuse & Alcoholism, Division of Biometry and Epidemiology, Alcohol
Epidemiologic Data System, Rockville, MD).
Yamada, T., M. Kendix and T. Yamada (1993), "The impact of alcohol consumption and marijuana
use on high school graduation", National Bureau of Economic Research Working Paper No. 4497,
October.
Yano, K., G.G. Rhoads, and A. Kagan. (1977),"Coffee, alcohol and risk of coronary heart disease
among Japanese men living in Hawaii", New England Journal of Medicine 297:405-409.
Zarkin G.A. et al. (1998), "Alcohol use and wages: New results from the National Household Survey
on Drug Abuse", Journal of Health Economics 17(1):53-68.
68
Figure 1: A Causal Model of Drinking and its Consequences
Earnings
Previous
Drinking,
Prices
Current
Drinking
Productivity
Human Capital
Health Status
69
Table 1: Economic Studies of Alcohol-Related Traffic Mortality
Authors
Data
Outcome
Policy
Result
Cook and
Tauchen (1984)
State-year
Highway Fatality
rates
MDA
-*
Males (1986)
State-year panel,
1975-83
Fatality rates
MDA
0
DuMouchel et al..
(1987)
State-year panel,
1975-84
Fatalities, age 1624
MDA
-*
Saffer and
Grossman (1987)
State-year panel,
1975-81
Fatality rates, by
age (youth)
Beer taxes
MDA
-*
-*
Chaloupka,
Saffer, and
Grossman (1993)
State-year panel,
1982-88
Fatality rates:
total, night, &
youth
Beer tax rates
MDA
Laws, Fines
Kenkel (1993)
1985 National
Health Interview
Survey
Self reports of
drinking & driving,
last 12 months
Price
MDA
knowledge
Monopoly
Deterrents1
Sloan et al.
(1995)
Behavioral Risk
Factor Survey
1984-90
Self reports of
drinking & driving,
last 30 days
Price
MDA
Deterrents
Liability rules
Ruhm (1996)
State-year panel
1982-88
Total & night
fatality rates
Beer tax rate
MDA
Laws, Fines, &
Penalties
-*
-*
mixed
-* (youths)
-*
-*
+*
-* (mostly)
-*
-*
mostly negative &
insignificant
-*
-* (18-20 only)
Mostly
insignificant
*Statistically significant effect at the p,.10 level.
1. Kenkel's results are shown for a heavy drinking equation. Heavy drinking significant increases drunk
driving in the second equation of his recursive framework.
70
Table 2: Drinking status and Medical Care
Former Drinker
Abstainer
Heavy Drinker1
HIE Outpatient Care
+*
+*
0
NHIS Outpatient Care
-*
0
+*
HIE Inpatient care
+*
+*
0
NHIS Inpatient Care
+*
+*
0
Source: Adapted from Manning et al. (1991) Tables 5-7 and 5-9. See text for discussion.
*Statistically significant effect at the p<.10 level.
1. The authors include the log of monthly ethanol consumption and the square of this variable in their
regressions. For the regressions on HIE data, the estimated coefficients are not in any case discernibly
different from zero. For the NHIS data the authors report that for the regression on outpatient data that
log consumption is significantly negative and the squared term is significantly positive.
71
Notes
1
See NTC Publications (1996).
2
See National Household Survey on Drug Abuse (1996).
3
For example, beginning during World War I (and ending in the mid 1950s) Sweden maintained a
complex rationing system in which citizens committees determined how much spirits each adult member
of the community could purchase based on such factors as his or her age, family and social
responsibilities, and reputation (Norström 1987).
4
See Grossman (1972).
5
Estimation using micro data must also reflect this feature. Two approaches have been suggested for
data such as these: The Tobit estimator, and the two-part estimator of Duan et al. (1984). The two part
estimator appears to be generally more preferable on a statistical basis.
6
In 31 out of 38 demand studies tabled in Edwards et al. (1994) that included estimated results for
both beer and spirits, the former was less price-elastic than the latter. A thorough review of
econometric studies using data for the United States (Leung and Phelps 1993) provides additional
confirmation.
7
NLSY's subsequent surveys continued to be at the respondent's home, but that home was
increasingly likely to be the respondent's than his or her parents'. The MTF's followup surveys, unlike
the initial survey, were mailed to the respondents' home. See also Hoyt and Chaloupka (1994).
8
Rare exceptions are Boland and Roizen (1973) and Popham and Schmidt (1981). Both studies
compare self-reports to sales data. The former study reported that heavier purchasers were more likely
to give accurate information to an interviewer. The latter reached the opposite conclusion, but based on
a somewhat faulty analysis of their data. They compare the distribution across buyers of the number of
72
bottles actually purchased to the distribution of the number of bottles reported purchased in the survey.
Both distributions are skewed to the right, but the mean of the self-reports is less than the mean of the
purchase records. The authors calculate the percentage of the population in each of a series of intervals
defined by the number of bottles. They find that the ratio (sales to reported purchases) of the
corresponding percentages from the two distributions increases with the number of bottles, and offer
that as evidence that underreporting is higher at higher quantities of purchases. But in fact this pattern is
compatible with a circumstance in which every drinker underreported by the same percentage.
9
At a more fundamental level, the consumption of alcohol is subject to a wide range of cultural
influences, including religious strictures, holiday traditions, popular entertainment, and other sources of
alcohol-related norms. These may provide another, more slow moving, feedback effect, responding
and amplifying trends in drinking.
10
Of course this result cannot hold over the entire range of possible prevalence levels, since it would rule
out the existence of an equilibrium value for group prevalence of drinking.
11
There are exceptions, however. Cook and Tauchen (1984) and DuMouchel et al. (1987) do not
control for beverage price changes. Chaloupka et al. (1993) substitute demographic characteristics for
state fixed effects.
12
Miron's (1997) econometric results lead him to challenge the belief that Prohibition had a large
effect on cirrhosis rates in the U.S. (Warburton 1932). A look at the data does suggest that the decline
in cirrhosis occurred before Prohibition: The age-adjusted cirrhosis mortality rate fell from 17.0 in 1911
to 8.9 in 1920 and remained at about that level through Prohibition and long after (DeBakey et al.,
1995). On the other hand, the influence of the Temperance movement was felt long before the 18th
73
Amendment. A number of states adopted prohibition before it became the nation's law, and these and
other restrictive measures may account for the early decline, which Prohibition then sustained.
13
See Moore (1996) for a report of similar findings with a somewhat different specification.
14
As Manning et al. (1991) notes, controlling for differences in health status diminishes the effects of
abstinence and cessation by one-fourth to one-half. The differences remain significant.
15
16
For reviews of this literature see Shaper (1990) and Lands and Zakhari (1991).
In fact there is a considerable social-science literature on the ways in which the job environment may
encourage or discourage drinking. Some occupations have long been associated with heavy drinking,
including those in which alcohol is readily available (bartenders, brewers) and in which workers are
often unsupervised (salespeople, farmworkers) (Trice and Sonnenstuhl 1988; Harford and Brooks
1992). The U.S. military was a particularly "wet" environment fueled by tax-free alcohol and heavydrinking traditions, until a more stringent set of policies on drinking and drugs was introduced in the early
1980s (Bray, Marsden, Herbold, and Peterson 1992).
17
It should be noted that in the United States, by one estimate 20 percent of all alcoholic beverage
sales are to businesses (Sammartino 1990, 76). No information is available on the incidence of this
portion of sales.
18
Lyon and Schwab (1995) suggest that the cross-section relationship between tax and income may
in part reflect differences with respect to location over the life cycle. But their calculations for the
alcohol excise taxes suggests that these taxes are highly regressive in a life-cycle framework as well.
19
Miller and Blincoe (1994) redo the estimate of motor-vehicle accidents to include nonfatal injuries.
They estimate that the external cost of drinking is $.63 per ounce just to account for these accidents.
74
20
Interestingly, in the 19th century "dram-shop" liability provided the wife of an alcoholic a cause of
action against the saloon where her husband was drinking away the household means of sustenance.
21
A can of beer, a glass of wine, and a shot of spirits all contain approximately the same amount of
ethanol, but tend to be taxed quite differently; in the United States, for example, the federal excise tax
on a shot of spirits exceeds the tax on a can of beer by a factor of two and on a glass of wine by a
factor of three. A number of commentators have advocated that tax rates be made uniform, the same
per unit of ethanol, regardless of the type of beverage (CSPI 1990). But the arguments in support of
this differentiation, based on claims about both regressivity and external cost, are not supported by
available evidence (Cook and Moore, 1993); in particular, there is little difference in regressivity, and
there is no basis for claiming that beer is the “drink of moderation.” For another perspective, see Saffer
(1994).
22
The violator may be judgment proof, in the sense of not being able to pay a fine as large as the cost
to the victim and society (Shavell 1986). Imposing other forms of punishment is socially costly. And if
the probability of being apprehended in less than one, the punishment must be greater than the harm to
preserve an appropriate deterrent. For a similar argument about gun control, see Cook and Leitzel
(1996).
75